Antenna array optimization method and system, terminal equipment and medium

Abstract

The invention provides an antenna array optimization method and system, terminal equipment and a medium. The method comprises the following steps: acquiring training data; training a pre-constructed antenna array optimization model by using the training data to obtain a trained antenna array optimization model; acquiring second position information of a to-be-optimized antenna array in the research area, and inputting the second position information into the trained antenna array optimization model to obtain an antenna array optimization strategy of the to-be-optimized antenna array; and optimizing the to-be-optimized antenna array according to the antenna array optimization strategy of the to-be-optimized antenna array. According to the invention, the optimization effect of the antenna array can be enhanced, and the UV coverage rate of the antenna array is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of microwave remote sensing and detection, and specifically relates to an antenna array optimization method, system, terminal device and medium. Background Technique

[0002] The synthetic aperture radiometer system (ASR) is a passive microwave remote sensing technology widely used in high-resolution imaging. This system works by multiple distributed antenna arrays collaborating to simulate an equivalent large-aperture antenna to obtain the radiation information of the target scene. The spatial frequency domain sampling coverage rate (UV coverage rate), as a key indicator for evaluating the spatial frequency domain sampling efficiency of the synthetic aperture radiometer system, directly reflects the sampling sufficiency of the system in the spatial frequency domain. The level of its value directly affects the system performance and the imaging quality of the target area, and directly determines the effects of subsequent target recognition and other applications.

[0003] The arrangement of the antenna array in the synthetic aperture radiometer system directly determines the UV coverage rate of the system. Due to the differences in the spacing and relative positions between antennas, different baseline vectors (i.e., the distance vectors formed between two antennas) will be generated. And these baseline vectors respectively correspond to different sampling points on the UV plane. A reasonable antenna array arrangement can form a uniform, dense and complementary set of sampling points on the UV plane, thereby improving the UV coverage rate; on the contrary, an unreasonable arrangement may cause sparse, missing or overly concentrated sampling points in some areas, resulting in a low UV coverage rate, and then seriously affecting the overall performance of the synthetic aperture radiometer system. Therefore, it is very important to optimize the array arrangement of the synthetic aperture radiometer to obtain the highest possible UV coverage rate under the same number of antennas.

[0004] Currently, traditional antenna array optimization methods mainly include optimization algorithms based on mathematical models and some empirical manual tuning methods, and there is still much room for improvement in terms of the arrangement effect of antennas and the reduction of computational complexity. Generally speaking, for common value strategy reinforcement learning algorithms, most application scenarios adopt a single-action decision-making mode, that is, at each time step, the agent can only select one action. This one-to-one decision-making method is usually applicable to scenarios with a small number of actions or independent actions. However, this mode has many technical bottlenecks that are difficult to break through when facing complex antenna array optimization problems.

[0005] From the perspective of system complexity, the optimization of antenna arrays involves the coordinated actions of multiple antennas. The optimization objectives need to consider the positions and movement directions of multiple antennas. The change in the position of each antenna not only affects its own signal reception but also changes the distribution of sampling points in the UV plane through the baseline vector, thereby affecting the UV coverage rate of the entire system. If the traditional one-to-one decision-making method is adopted, only one antenna action can be adjusted each time, which will not only affect one sampling baseline but also all the baselines corresponding to the antenna pairs formed by this antenna. Therefore, the traditional one-to-one optimization cannot consider the mutual influence and coordination relationship between the actions of multiple antennas, resulting in extremely low optimization efficiency and difficulty in finding the optimal antenna array layout scheme within a reasonable time.

[0006] From the analysis of the algorithm decision-making mechanism, in the optimization of antenna arrays, when multiple antennas act simultaneously, the state changes generated by different action combinations vary greatly. Traditional algorithms are difficult to effectively evaluate and learn these complex state changes, resulting in poor optimization effects of antenna arrays and low UV coverage rates. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide an antenna array optimization method, system, terminal device and medium to enhance the optimization effect of antenna arrays and improve the UV coverage rate of antenna arrays.

[0008] In a first aspect, the present invention provides an antenna array optimization method, which includes the following steps:

[0009] Obtain training data; the training data includes the first position information corresponding to multiple antennas;

[0010] Use the training data to train a pre-constructed antenna array optimization model to obtain a trained antenna array optimization model; wherein, the antenna array optimization model is constructed based on a residual skip connection network. The antenna array optimization model takes the position information of multiple antennas at the current moment as input, and takes the position change information of each antenna in the multiple antennas as the antenna array optimization strategy and outputs it. The antenna array optimization strategy includes multiple actions, and a reward function is constructed based on the spatial frequency domain sampling coverage rate to train the antenna array optimization model, and the antenna array optimization strategy is determined through action value.

[0011] Obtain the second position information of the antenna array to be optimized in the research area, and input the second position information into the trained antenna array optimization model to obtain the antenna array optimization strategy of the antenna array to be optimized;

[0012] Optimize the antenna array to be optimized according to the antenna array optimization strategy of the antenna array to be optimized.

[0013] Optionally, the constructed antenna array optimization model is trained using training data to obtain a trained antenna array optimization model, including:

[0014] Step I, calculate the first spatial frequency domain sampling coverage rate corresponding to the first position information, and initialize the antenna array optimization model parameters to obtain an initial antenna array optimization model;

[0015] Step II, input the first position information into the initial antenna array optimization model to obtain an initial antenna array optimization strategy output by the initial antenna array optimization model; the antenna array optimization strategy includes the moving direction and moving step size of each antenna in multiple antennas; the moving step size is preset;

[0016] Step III, adjust the positions of each antenna according to the initial antenna array optimization strategy to obtain the third position information corresponding to multiple antennas, and calculate the second spatial frequency domain sampling coverage rate corresponding to the third position information;

[0017] Step IV, calculate the loss value of the initial antenna array optimization model according to the first spatial frequency domain sampling coverage rate and the second spatial frequency domain sampling coverage rate, and perform backpropagation on the initial antenna array optimization model according to the loss value until the loss value of the antenna array optimization model is less than a preset loss threshold to obtain a trained antenna array optimization model.

[0018] Optionally, the expression of the reward function is:

[0019] reward = part1 + part2

[0020] Or

[0021]

[0022] ΔC = coverage2 - coverage1

[0023] wherein, reward represents the reward value, coverage1 represents the spatial frequency domain sampling coverage rate of multiple antennas before optimization, coverage2 represents the spatial frequency domain sampling coverage rate of multiple antennas after optimization, ΔC represents the difference in spatial frequency domain sampling coverage rate before and after optimization, when ΔC≥0, when ΔC < 0, part1 represents the dynamic reward based on the change in spatial frequency domain sampling coverage rate, and part2 represents the reward based on the spatial frequency domain sampling coverage rate after optimization.

[0024] Optionally, the calculation expression of the loss value is:

[0025] loss = (q_value - q_target) 2

[0026]

[0027] q_target = reward + γ * max_Value

[0028]

[0029] Among them, loss represents the loss value, q_value represents the average action value corresponding to the initial antenna array optimization strategy, reward m represents the action value corresponding to the m-th action in the initial antenna array optimization strategy, m = 1, 2,..., M, M represents the total number of actions, γ represents the proportionality coefficient, max_Value represents the average action value corresponding to the new antenna array optimization strategy obtained by inputting the position information of multiple optimized antennas into the antenna array optimization model, reward' m represents the action value corresponding to the m-th action in the new antenna array optimization strategy.

[0030] Optionally, the antenna array optimization model includes a first convolutional layer, a second convolutional layer, a skip connection layer, a first fully connected layer, and a second fully connected layer;

[0031] The input end of the first convolutional layer and the input end of the skip connection layer receive the position information of the antenna array. The output end of the first convolutional layer is connected to the input end of the second convolutional layer. The input end of the first fully connected layer is connected to the output end of the second convolutional layer and the output end of the skip connection layer. The output end of the first fully connected layer is connected to the input end of the second fully connected layer. The output end of the second fully connected layer outputs the antenna array optimization strategy.

[0032] Optionally, in the trained antenna array optimization model, the convolutional kernel size of the first convolutional layer is set to 3, the stride is set to 1, the padding is set to 1, and the number of feature channels is set to 16;

[0033] The convolutional kernel size of the second convolutional layer is set to 3, the stride is set to 1, the padding is set to 1, and the number of feature channels is set to 32;

[0034] The skip connection layer is a one-dimensional convolutional layer, the convolutional kernel size is 1, and the number of feature channels is set to 32.

[0035] Optionally, the action is used to describe the moving direction of the antenna;

[0036] Determining the antenna array optimization strategy through the action value includes:

[0037] For each antenna respectively, obtain multiple actions corresponding to the antenna, calculate the action value corresponding to each action among the multiple actions corresponding to the antenna respectively, and determine the moving direction of the antenna in the antenna array optimization strategy as the action with the maximum action value;

[0038] According to the moving directions and moving step lengths of all antennas, obtain the antenna array optimization strategy.

[0039] In a second aspect, the present invention discloses an antenna array optimization system, including:

[0040] A data acquisition module, configured to acquire training data; the training data includes first position information corresponding to multiple antennas;

[0041] A model training module, configured to train a pre-constructed antenna array optimization model using the training data to obtain a trained antenna array optimization model; wherein, the antenna array optimization model is constructed based on a residual skip connection network, the antenna array optimization model takes the position information of multiple antennas at the current moment as input, takes the position change information of each antenna among the multiple antennas as the antenna array optimization strategy and outputs, the antenna array optimization strategy includes multiple actions, and constructs a reward function based on the spatial frequency domain sampling coverage rate, trains the antenna array optimization model with the reward function, and determines the antenna array optimization strategy through action values;

[0042] A strategy acquisition module, configured to acquire second position information of the antenna array to be optimized in the research area, input the second position information into the trained antenna array optimization model, and obtain the antenna array optimization strategy of the antenna array to be optimized;

[0043] An array optimization module, configured to optimize the antenna array to be optimized according to the antenna array optimization strategy of the antenna array to be optimized.

[0044] In a third aspect, the present invention discloses a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the above method is implemented.

[0045] In a fourth aspect, the present invention discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above method is implemented.

[0046] The beneficial effects of the present invention are:

[0047] The antenna array optimization method disclosed in the present invention constructs an antenna array optimization model that breaks through the limitation of one-to-one correspondence between decisions and actions in traditional methods. The constructed antenna array optimization strategy includes multiple actions, which can adjust all antennas for antenna array optimization simultaneously, avoiding the influence of single antenna actions on the overall sampling baseline, facilitating enhancing the antenna array optimization effect, and improving the UV coverage rate of the antenna array. The reward function constructed based on the spatial frequency domain sampling coverage rate can accurately screen out the actions that can most improve the UV coverage rate, thereby significantly enhancing the UV coverage rate of the antenna array. Description of the Drawings

[0048] Figure 1 is the imaging principle diagram of a synthetic aperture radiometer;

[0049] Figure 2 is the flowchart of the antenna array optimization method in one embodiment of the present invention;

[0050] Figure 3 is the structural diagram of the antenna array optimization model in one embodiment of the present invention;

[0051] Figure 4 is the comparison chart of the UV coverage rate between the antenna array optimization method provided by the present invention and the traditional method in one embodiment of the present invention;

[0052] Figure 5 is the structural diagram of the antenna array optimization system in one embodiment of the present invention;

[0053] Figure 6 is the structural diagram of the terminal device in one embodiment of the present invention. Detailed Embodiments

[0054] Aiming at the technical problems of poor antenna array optimization effect and low UV coverage rate of traditional antenna array optimization methods, the present invention discloses an antenna array optimization method, system, terminal device, and medium. Among them, the antenna array optimization model constructed by this method breaks through the limitation of one-to-one correspondence between decisions and actions in traditional methods. The constructed antenna array optimization strategy includes multiple actions, which can adjust all antennas for antenna array optimization simultaneously, avoiding the influence of single antenna actions on the overall sampling baseline, facilitating enhancing the antenna array optimization effect, and improving the UV coverage rate of the antenna array. The reward function constructed based on the spatial frequency domain sampling coverage rate can accurately screen out the actions that can most improve the UV coverage rate, thereby significantly enhancing the UV coverage rate of the antenna array.

[0055] For ease of understanding, the principle of synthetic aperture radiometer imaging is first described as Figure 1 shown.

[0056] On the OXY plane, any two points (xi , y i , z i , ) and (x k , y k , z k , ) positions are respectively provided with two receiving antennas (antenna i and antenna k), and the two receiving antennas respectively receive the radiation from the radiation source d σ (x σ , y σ , z σ , ) and transmit the received signals to two receiving channels (the boxes in the figure), and then perform complex correlation on the channel output signals. These two antennas form an antenna pair, and the spatial distance d between the antenna pair y and the direction of the distance are called the baseline vector, simply referred to as the baseline. The complex correlation output V of this antenna pair is called the visibility. Since the visibility is a function of the baseline, it is called the visibility function, and finally the brightness temperature image can be obtained through reconstruction.

[0057] The visibility function is related to the direction and length of the baseline, and the baseline depends on the relative positions of the antennas. Each baseline vector formed by a pair of antennas corresponds to a sampling point on the UV plane. When there are N antennas forming an antenna array, all these baselines will form a sampling distribution on the UV plane, that is, UV coverage. The higher the UV coverage rate, the more sufficient the sampling in the spatial frequency domain, the more signal details can be captured, and the overall performance of the synthetic aperture radiometer system is higher, thus making the imaging quality higher.

[0058] Referring to the above description, the antenna array optimization method disclosed in the present invention will be described below. As Figure 2 shown, the antenna array optimization method includes steps 21 to 24.

[0059] Step 21, obtain training data.

[0060] In the embodiments of the present invention, the above training data includes the first position information corresponding to multiple antennas. Exemplarily, the two-dimensional position coordinates of multiple antennas in a certain area can be obtained from a public dataset, and the first position information can be expressed as State = {(x1, y1), (x2, y2),..., (x n , y n )}; where, (x n , y n ) represents the coordinate position of the nth antenna in the two-dimensional plane in this area. For the convenience of the subsequent reinforcement learning process, the present invention takes the position information of multiple antennas as the system state.

[0061] Step 22, use the training data to train a pre-constructed antenna array optimization model to obtain a trained antenna array optimization model.

[0062] Among them, the antenna array optimization model is constructed based on a residual jump connection network. The antenna array optimization model takes the position information of multiple antennas at the current moment as input, and takes the position change information of each antenna among the multiple antennas as the antenna array optimization strategy and outputs it. The antenna array optimization strategy includes multiple actions, and a reward function is constructed based on the spatial frequency domain sampling coverage rate. The antenna array optimization model is trained with the reward function, and the antenna array optimization strategy is determined through action values.

[0063] In an embodiment of the present invention, the expression of the reward function is:

[0064] reward = part1 + part2

[0065] Or

[0066]

[0067] ΔC = coverage2 - coverage1

[0068] Among them, reward represents the reward value, coverage1 represents the spatial frequency domain sampling coverage rate of multiple antennas before optimization, coverage2 represents the spatial frequency domain sampling coverage rate of multiple antennas after optimization, ΔC represents the difference in the spatial frequency domain sampling coverage rate before and after optimization. When ΔC≥0, When ΔC < 0, part1 represents the dynamic reward based on the change in the spatial frequency domain sampling coverage rate, and part2 represents the reward based on the spatial frequency domain sampling coverage rate after optimization.

[0069] The calculation expression of the loss value is:

[0070] loss = (q_value - q_target) 2

[0071]

[0072] q_target = reward + γ * max_Value

[0073]

[0074] Among them, loss represents the loss value, q_value represents the average action value corresponding to the initial antenna array optimization strategy, reward mDenote the action value corresponding to the m-th action in the initial antenna array optimization strategy, where m = 1, 2, ..., M, M represents the total number of actions, γ represents the proportionality coefficient, max_Value represents the average action value corresponding to the new antenna array optimization strategy obtained by inputting the position information of multiple optimized antennas into the antenna array optimization model, and reward′ m Denote the action value corresponding to the m-th action in the new antenna array optimization strategy.

[0075] The model structure of the antenna array optimization model disclosed in the present invention will be described below.

[0076] Specifically, as Figure 3 shown, the antenna array optimization model includes a first convolutional layer 301, a second convolutional layer 302, a skip connection layer 303, a first fully connected layer 304, and a second fully connected layer 305. Among them, the input end of the first convolutional layer 301 and the input end of the skip connection layer 303 receive the position information of the antenna array. The output end of the first convolutional layer 301 is connected to the input end of the second convolutional layer 302. The input end of the first fully connected layer 304 is connected to the output end of the second convolutional layer 302 and the output end of the skip connection layer 303. The output end of the first fully connected layer 304 is connected to the input end of the second fully connected layer 305, and the output end of the second fully connected layer 305 outputs the antenna array optimization strategy.

[0077] The training process of the antenna array optimization model in the embodiments of the present invention will be described below, specifically including steps I to IV:

[0078] Step I, calculate the first spatial frequency domain sampling coverage rate corresponding to the first position information, and initialize the parameters of the antenna array optimization model to obtain the initial antenna array optimization model.

[0079] The spatial frequency domain sampling coverage rate can be calculated by common methods. Exemplarily, when the coverage area is a circular area, the UV coverage rate can be obtained through the calculation formula

[0080]

[0081] ; where u max represents the maximum spatial frequency, and λ represents the wavelength of the radiation signal.

[0082] In other embodiments of the present application, the coverage area may also be other shapes. Since the research on the UV coverage rate is relatively mature, the UV coverage rate can be calculated by common UV coverage rate calculation methods, which will not be elaborated here.

[0083] Step II, input the first position information into the initial antenna array optimization model to obtain the initial antenna array optimization strategy output by the initial antenna array optimization model.

[0084] The antenna array optimization strategy includes the moving direction and moving step of each antenna among multiple antennas. The moving step is preset.

[0085] Exemplarily, the moving directions include four directions: east, south, west, and north, and the moving step can be set to 0.2 meters.

[0086] It should be noted that in the embodiments of the present invention, the set of actions composed of all antennas is used as the action space of reinforcement learning, and each action is used to describe the moving direction of a certain antenna. The process of determining the antenna array optimization strategy based on the action value is described below, specifically including step a to step b.

[0087] In step a, for each antenna, obtain multiple actions corresponding to the antenna, calculate the action value corresponding to each action among the multiple actions corresponding to the antenna, and determine the moving direction of the antenna in the antenna array optimization strategy as the action with the maximum action value.

[0088] In step b, according to the moving directions and moving steps of all antennas, obtain the antenna array optimization strategy.

[0089] In step III, adjust the positions of each antenna according to the initial antenna array optimization strategy to obtain the third position information corresponding to multiple antennas, and calculate the second spatial frequency domain sampling coverage rate corresponding to the third position information.

[0090] In step IV, calculate the loss value of the initial antenna array optimization model according to the first spatial frequency domain sampling coverage rate and the second spatial frequency domain sampling coverage rate, and perform backpropagation on the initial antenna array optimization model according to the loss value until the loss value of the antenna array optimization model is less than the preset loss threshold to obtain the trained antenna array optimization model.

[0091] Exemplarily, in the embodiments of the present invention, the model parameters of the trained antenna array optimization model are as follows:

[0092] The convolution kernel size of the first convolutional layer is set to 3, the stride is set to 1, the padding is set to 1, and the number of feature channels is set to 16;

[0093] The convolution kernel size of the second convolutional layer is set to 3, the stride is set to 1, the padding is set to 1, and the number of feature channels is set to 32;

[0094] The skip connection layer is a one-dimensional convolutional layer, the convolution kernel size is 1, and the number of feature channels is set to 32.

[0095] In step 23, obtain the second position information of the antenna array to be optimized within the research area, and input the second position information into the trained antenna array optimization model to obtain the antenna array optimization strategy of the antenna array to be optimized.

[0096] Step 24: Optimize the antenna array to be optimized according to the antenna array optimization strategy of the antenna array to be optimized.

[0097] Specifically, by adjusting the positions of the antennas in the antenna array to be optimized according to the position change information corresponding to the antenna array to be optimized in the antenna array optimization strategy, the maximum UV coverage rate can be achieved.

[0098] The antenna array optimization method provided by the present invention will be further described below in conjunction with embodiments.

[0099] In this embodiment, the traditional antenna array optimization method (one decision corresponds to one action, that is, one antenna is moved at a time) and the antenna array optimization method provided by the present invention (all antennas are moved at a time) are used for comparison. Specifically, in a 20*20 area, 4 antennas are fixed at the four top corners of the area, and then the positions of 16 movable antennas in the area are optimized. The results are as Figure 4 shown. It is not difficult to see that compared with the traditional antenna array optimization method, the antenna array optimization method provided by the present invention shows a rapid upward trend in the UV coverage rate within the same number of training rounds. In the initial stage of training, its growth rate far exceeds the case of moving one antenna; as the number of training rounds increases, this gap becomes more obvious. At the end of 1000 rounds of training, the UV coverage rate of moving 16 antennas reaches a relatively high value, far exceeding the level that can be achieved by moving one antenna. This comparison fully shows that when optimizing the antenna array by the traditional antenna array optimization method, due to the inability to consider the collaborative relationship between multiple antenna actions, the optimization efficiency is low and the improvement effect of the UV coverage rate is not good. However, the antenna array optimization method provided by the present invention can take into account the mutual influence and collaborative relationship between multiple antenna actions, effectively break through this limitation, and significantly improve the UV coverage rate through the collaborative movement of multiple antennas, providing a more efficient and better solution for antenna array optimization.

[0100] In summary, for the antenna array optimization method disclosed by the present invention, the constructed antenna array optimization model breaks through the limitation of one-to-one correspondence between decisions and actions in the traditional method. The constructed antenna array optimization strategy includes multiple actions, which can simultaneously adjust all the antennas for antenna array optimization, avoiding the influence of single antenna actions on the overall sampling baseline, being beneficial to enhancing the antenna array optimization effect and improving the UV coverage rate of the antenna array; the reward function constructed based on the spatial frequency domain sampling coverage rate can accurately screen out the actions that can most improve the UV coverage rate, thereby significantly improving the UV coverage rate of the antenna array.

[0101] The antenna array optimization system disclosed by the present invention will be described below.

[0102] As Figure 5As shown in the figure, the antenna array optimization system 500 includes:

[0103] A data acquisition module 501 for acquiring training data; the training data includes first position information corresponding to multiple antennas;

[0104] A model training module 502 for training a pre-constructed antenna array optimization model using the training data to obtain a trained antenna array optimization model; wherein, the antenna array optimization model is constructed based on a residual skip connection network, and the antenna array optimization model takes the position information of multiple antennas at the current moment as input, takes the position change information of each antenna among the multiple antennas as the antenna array optimization strategy and outputs it. The antenna array optimization strategy includes multiple actions, and a reward function is constructed based on the spatial frequency domain sampling coverage rate to train the antenna array optimization model, and the antenna array optimization strategy is determined through action values;

[0105] A strategy acquisition module 503 for acquiring second position information of the antenna array to be optimized within the research area, inputting the second position information into the trained antenna array optimization model, and obtaining the antenna array optimization strategy of the antenna array to be optimized;

[0106] An array optimization module 504 for optimizing the antenna array to be optimized according to the antenna array optimization strategy of the antenna array to be optimized.

[0107] It should be noted that for the information interaction, execution process, etc. between the above modules, since they are based on the same concept as the method embodiment of the present application, their specific functions and the technical effects brought about can be specifically referred to in the method embodiment part, and will not be elaborated here. Those skilled in the art can clearly understand that for the convenience and conciseness of description, only the above division of each functional unit and module is used as an example for illustration. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of the present application. The specific working processes of the units and modules in the above system can refer to the corresponding processes in the foregoing method embodiment and will not be elaborated here.

[0108] As Figure 6 shown, an embodiment of the present invention provides a terminal device. As Figure 6 shown, the terminal device D10 of this embodiment includes: at least one processor D100 (Figure 6 only shows one processor), a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100. When the processor D100 executes the computer program D102, the steps in any of the above method embodiments are implemented.

[0109] Specifically, when the processor D100 executes the computer program D102, it obtains training data; uses the training data to train a pre-constructed antenna array optimization model to obtain a trained antenna array optimization model; obtains second position information of the antenna array to be optimized within the research area, inputs the second position information into the trained antenna array optimization model to obtain an antenna array optimization strategy for the antenna array to be optimized; and optimizes the antenna array to be optimized according to the antenna array optimization strategy of the antenna array to be optimized. Among them, the constructed antenna array optimization model breaks through the limitation of the one-to-one correspondence between decisions and actions in traditional methods. The constructed antenna array optimization strategy includes multiple actions, which can adjust all antennas for antenna array optimization simultaneously, avoiding the influence of single antenna actions on the overall sampling baseline, facilitating enhancing the antenna array optimization effect and improving the UV coverage rate of the antenna array; the reward function constructed based on the spatial frequency domain sampling coverage rate can accurately screen out the actions that can most improve the UV coverage rate, thus significantly improving the UV coverage rate of the antenna array.

[0110] The so-called processor D100 may be a central processing unit (CPU, Central Processing Unit), and this processor D100 may also be other general-purpose processors, digital signal processors (DSP, Digital Signal Processor), application specific integrated circuits (ASIC, Application Specific Integrated Circuit), field-programmable gate arrays (FPGA, Field-Programmable Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0111] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as the hard disk or memory of the terminal device D10. In some other embodiments, the memory D101 may also be an external storage device of the terminal device D10, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc., equipped on the terminal device D10. Further, the memory D101 may also include both the internal storage unit of the terminal device D10 and the external storage device. The memory D101 is used to store an operating system, application programs, a BootLoader, data, and other programs, such as the program code of the computer program. The memory D101 may also be used to temporarily store data that has been output or is to be output.

[0112] An embodiment of the present application also provides a computer-readable storage medium storing a computer program, which when executed by a processor, can implement the steps in the above-mentioned method embodiments.

[0113] An embodiment of the present application provides a computer program product, which when running on a terminal device, enables the terminal device to execute and implement the steps in the above-mentioned method embodiments.

[0114] Those of ordinary skill in the art should understand that: The discussion of any of the above embodiments is only exemplary and is not intended to imply that the protection scope of the present application is limited to these examples; Under the concept of the present application, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of one or more embodiments of the present application as described above, which are not provided in detail for the sake of brevity.

[0115] One or more embodiments of the present application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of the present application shall be included in the protection scope of the present application.

Claims

1. An antenna array optimization method, characterized in that: include: Get training data; The training data includes first position information corresponding to a plurality of antennas; The pre-constructed antenna array optimization model is trained using the training data to obtain a trained antenna array optimization model; wherein the antenna array optimization model is constructed based on a residual skip connection network, the antenna array optimization model uses the position information of multiple antennas at the current moment as input, and uses the position change information of each antenna in the multiple antennas as an antenna array optimization strategy and outputs it, the antenna array optimization strategy includes multiple actions, and a reward function is constructed based on the spatial frequency domain sampling coverage, the antenna array optimization model is trained with the reward function, and the antenna array optimization strategy is determined by the action value; Acquire second position information of the antenna array to be optimized in the study area, input the second position information into the trained antenna array optimization model, and obtain an antenna array optimization strategy for the antenna array to be optimized; The antenna array to be optimized is optimized according to the antenna array optimization strategy of the antenna array to be optimized.

2. The antenna array optimization method according to claim 1, characterized in that: The step of training the constructed antenna array optimization model using the training data to obtain a trained antenna array optimization model includes: Step I, calculating the first spatial frequency domain sampling coverage corresponding to the first position information, and initializing antenna array optimization model parameters to obtain an initial antenna array optimization model; Step II, inputting the first position information into the initial antenna array optimization model to obtain an initial antenna array optimization strategy output by the initial antenna array optimization model; the antenna array optimization strategy includes a moving direction and a moving step length of each antenna in the multiple antennas; the moving step length is preset; Step III, adjusting the position of each antenna according to the initial antenna array optimization strategy, obtaining third position information corresponding to the multiple antennas, and calculating the second spatial frequency domain sampling coverage corresponding to the third position information; Step IV, calculate the loss value of the initial antenna array optimization model based on the first spatial frequency domain sampling coverage and the second spatial frequency domain sampling coverage, and back-propagate the initial antenna array optimization model based on the loss value until the loss value of the antenna array optimization model is less than a preset loss threshold, thereby obtaining the trained antenna array optimization model.

3. The antenna array optimization method according to claim 2, characterized in that: The expression of the reward function is: reward = part1 + part2 or ΔC=coverage2-coverage1 Among them, reward represents the reward value, coverage1 represents the spatial frequency domain sampling coverage of multiple antennas before optimization, coverage2 represents the spatial frequency domain sampling coverage of multiple antennas after optimization, ΔC represents the difference in spatial frequency domain sampling coverage before and after optimization, and when ΔC ≥ 0, When ΔC<0, Part 1 represents the dynamic reward based on the change of spatial frequency domain sampling coverage, and part 2 represents the reward based on the optimized spatial frequency domain sampling coverage.

4. The antenna array optimization method according to claim 3, characterized in that: The calculation expression of the loss value is: loss=(q_value-q_target) 2 q_target=reward+γ*max_Value Among them, loss represents the loss value, q_value represents the average action value corresponding to the initial antenna array optimization strategy, and reward m represents the action value corresponding to the mth action in the initial antenna array optimization strategy, m = 1, 2, ..., M, M represents the total number of actions, γ represents the proportional coefficient, max_Value represents the average action value corresponding to the new antenna array optimization strategy obtained by inputting the position information of multiple optimized antennas into the antenna array optimization model, reward′ m Represents the action value corresponding to the mth action in the new antenna array optimization strategy.

5. The antenna array optimization method according to claim 4, characterized in that: The antenna array optimization model includes a first convolutional layer, a second convolutional layer, a skip connection layer, a first fully connected layer and a second fully connected layer; The input end of the first convolutional layer and the input end of the jump connection layer receive the position information of the antenna array, the output end of the first convolutional layer is connected to the input end of the second convolutional layer, the input end of the first fully connected layer is connected to the output end of the second convolutional layer and the output end of the jump connection layer, the output end of the first fully connected layer is connected to the input end of the second fully connected layer, and the output end of the second fully connected layer outputs the antenna array optimization strategy.

6. The antenna array optimization method according to claim 5, characterized in that: In the trained antenna array optimization model, the convolution kernel size of the first convolution layer is set to 3, the stride is set to 1, the padding is set to 1, and the number of feature channels is set to 16; The convolution kernel size of the second convolutional layer is set to 3, the stride is set to 1, the padding is set to 1, and the number of feature channels is set to 32; The jump connection layer is a one-dimensional convolution layer, the convolution kernel size is 1, and the number of feature channels is set to 32.

7. The antenna array optimization method according to claim 6, characterized in that: The action is used to describe the moving direction of the antenna; The step of determining the antenna array optimization strategy by action value includes: For each of the antennas, obtain multiple actions corresponding to the antenna, calculate the action value corresponding to each of the multiple actions corresponding to the antenna, and determine the action with the largest action value as the moving direction of the antenna in the antenna array optimization strategy; The antenna array optimization strategy is obtained according to the moving directions of all antennas and the moving step lengths.

8. An antenna array optimization system, characterized in that: include: A data acquisition module, used to acquire training data; The training data includes first position information corresponding to a plurality of antennas; A model training module, used to train a pre-constructed antenna array optimization model using the training data to obtain a trained antenna array optimization model; wherein the antenna array optimization model is constructed based on a residual skip connection network, the antenna array optimization model uses the position information of multiple antennas at the current moment as input, and uses the position change information of each antenna in the multiple antennas as an antenna array optimization strategy and outputs it, the antenna array optimization strategy includes multiple actions, and a reward function is constructed based on the spatial frequency domain sampling coverage, the antenna array optimization model is trained with the reward function, and the antenna array optimization strategy is determined by the action value; A strategy acquisition module, used to acquire second position information of the antenna array to be optimized in the study area, input the second position information into the trained antenna array optimization model, and obtain an antenna array optimization strategy for the antenna array to be optimized; The array optimization module is used to optimize the antenna array to be optimized according to the antenna array optimization strategy of the antenna array to be optimized.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images