Satellite-borne phase-controlled array digital-analog hybrid beamforming method, device and equipment

Abstract

The application relates to the technical field of satellite communication, and discloses a satellite-borne phased array digital-analog hybrid beamforming method, device and equipment. The method comprises the following steps: configuring a transmitting architecture of a satellite-borne phased array antenna and a double-time-scale communication frame structure; in the starting stage of each communication frame, extracting a main radio frequency subspace according to a channel of a starting time slot, generating an initial analog beamforming matrix and configuring the initial analog beamforming matrix to a phase shifter network; in each time slot in the communication frame, updating the main radio frequency subspace of a previous time slot by using incremental singular value decomposition, generating an analog beamforming matrix of the current time slot based on the updated main radio frequency subspace; in each time slot in the communication frame, generating a digital beamforming matrix of the current time slot based on a liquid neural network; and at the end of each communication frame, correcting parameters of the liquid neural network based on a loss function accumulated in the frame. The application can realize a light-weight digital-analog hybrid beamforming design, guarantee communication performance, and significantly reduce online operation burden.

Description

Technical Field

[0001] This application relates to the field of satellite communication technology, and in particular to a method, apparatus and equipment for satellite-borne phased array hybrid digital-analog beamforming. Background Technology

[0002] With the rapid development of low-Earth orbit satellite internet and high-throughput communication technologies, spaceborne communication systems are facing higher demands for spectral efficiency and multi-user access capabilities. Large-scale phased array technology, with its high gain and flexible beam pointing characteristics, has become a key means to improve satellite communication capacity. Given the strict limitations of spaceborne platforms in terms of power consumption, heat dissipation, and hardware cost, a hybrid analog-digital architecture with lower hardware complexity is typically used to implement beamforming design.

[0003] However, hybrid analog-digital beamforming design typically involves complex non-convex optimization problems. Traditional iterative algorithms suffer from high computational overhead and slow convergence speed, making it difficult to meet the stringent real-time requirements of spaceborne communication. Furthermore, the high-speed motion of low-Earth orbit satellites causes drastic dual dynamic changes in the channel: instantaneous rapid fading caused by multipath effects and spatial angular drift due to orbital motion. Existing technologies are mostly based on static or quasi-static channel assumptions, making it difficult to maintain precise narrow beam alignment when the channel changes rapidly, easily leading to communication interruptions. Although artificial intelligence technology has been introduced into beamforming design in recent years to reduce computational overhead, most existing deep learning solutions rely on massive offline training, with fixed model parameters and high computational demands. These solutions not only lack online adaptive capabilities to cope with continuously time-varying channels, but their high computing power and storage requirements are also difficult to adapt to the resource-constrained spaceborne processing units. Summary of the Invention

[0004] This application provides a method, apparatus, and device for spaceborne phased array hybrid digital-analog beamforming, which at least solves the technical problem in related technologies that it is difficult to achieve a good balance between computational overhead and beam tracking accuracy.

[0005] According to one aspect of the embodiments of this application, a spaceborne phased array hybrid digital-analog beamforming method is provided, comprising:

[0006] The system is configured with a satellite-borne phased array antenna transmission architecture and a dual-time-scale communication frame structure, wherein the frame structure discretizes the communication time axis into frame-level and time-slot-level scales.

[0007] At the beginning of each communication frame, the main radio frequency subspace is extracted according to the channel of the initial time slot, an initial simulated beamforming matrix is ​​generated and configured for the phase shifter network;

[0008] In each time slot within the communication frame, incremental singular value decomposition is used to update the main radio frequency subspace of the previous time slot, and the simulated beamforming matrix of the current time slot is generated based on the updated main radio frequency subspace.

[0009] In each time slot within the communication frame, a digital beamforming matrix for the current time slot is generated based on the state evolution of a liquid neural network;

[0010] At the end of each communication frame, the parameters of the liquid neural network are modified based on the accumulated loss function within that frame, so that the liquid neural network can adapt to channel changes.

[0011] In one implementation, the transmission architecture configured with a spaceborne phased array antenna and a dual-time-scale communication frame structure includes:

[0012] Initialize the downlink communication link configuration of the low-Earth orbit satellite and enable the phased array antenna and hybrid digital-analog beamforming architecture as the signal transmitter;

[0013] Configure the time-domain parameters of the communication link and construct a dual-time-scale frame structure. The frame structure discretizes the communication time axis into a frame-level scale and a time-slot-level scale. The frame-level scale is used to characterize the slow-varying characteristics of the channel, and the time-slot-level scale is used to characterize the fast-fading characteristics of the channel.

[0014] Based on the phased array antenna, the hybrid digital-analog beamforming architecture, and the dual-time-scale frame structure, the spectral efficiency optimization target of the system is calculated and established.

[0015] In one implementation, at the beginning stage of each communication frame, the main radio frequency subspace is extracted based on the channel of the initial time slot, an initial analog beamforming matrix is ​​generated and configured for the phase shifter network, including:

[0016] At the beginning of each communication frame, the instantaneous channel matrix is ​​acquired, and singular value decomposition is performed on the channel matrix to extract the basis matrix of the main radio frequency subspace.

[0017] An optimization problem is established based on the subspace approximation criterion, and the optimization problem is solved under constant mode constraints to obtain the initial simulated beamforming matrix.

[0018] The corresponding phase control commands are generated based on the initial simulated beamforming matrix and sent to the phase shifter network.

[0019] In one implementation, within each time slot of the communication frame, the main radio frequency subspace of the previous time slot is updated using incremental singular value decomposition. Based on the updated main radio frequency subspace, an analog beamforming matrix for the current time slot is generated, including:

[0020] When a new channel sample arrives, the new channel features are extracted based on the main radio frequency subspace of the previous time slot, and the first incremental matrix is ​​constructed.

[0021] Perform singular value decomposition on the first incremental matrix and update the main radio frequency subspace basis matrix of the previous time slot;

[0022] Based on the updated main radio frequency subspace basis matrix, the simulated beamforming matrix for the current time slot is constructed under constant mode constraints.

[0023] In one implementation, when a new channel sample arrives, based on the main radio frequency subspace of the previous time slot, the newly added channel features are extracted, and a first incremental matrix is ​​constructed, including:

[0024] When a new channel sample arrives, the channel projection and orthogonal residual components of the current time slot are calculated based on the main radio frequency subspace of the previous time slot.

[0025] The channel projection and orthogonal residual components are orthogonally normalized to obtain the new channel features;

[0026] Based on the newly added channel features, and by introducing a forgetting factor and combining historical singular value information, the first incremental matrix is ​​constructed.

[0027] In one implementation, within each time slot of the communication frame, a digital beamforming matrix for the current time slot is generated based on the state evolution of a liquid neural network, including:

[0028] A hot-start mechanism is used to initialize the current state of the liquid neural network, and the gradient of the instantaneous loss function is calculated based on the beamforming optimization variables of the previous time slot.

[0029] The gradient of the instantaneous loss function is input into the liquid neural network to drive the liquid neural network to perform continuous-time hidden state evolution;

[0030] Based on the evolved hidden state, output the second incremental matrix used to update the digital beam;

[0031] The digital beamforming basis matrix is ​​updated based on the second incremental matrix to generate the digital beamforming matrix for the current time slot.

[0032] In one implementation, at the end of each communication frame, the parameters of the liquid neural network are corrected based on the accumulated loss function within that frame, including:

[0033] At the end of each communication frame, the trainable parameters of the liquid neural network are corrected using the backpropagation algorithm based on the accumulated loss function of all time slots within that frame.

[0034] According to another aspect of the embodiments of this application, a spaceborne phased array hybrid digital-analog beamforming device is provided, comprising:

[0035] The communication link configuration module is used to configure the transmission architecture of the spaceborne phased array antenna and the dual-time-scale communication frame structure, wherein the frame structure discretizes the communication time axis into frame-level scale and time-slot-level scale.

[0036] The analog beam initialization module is used to extract the main radio frequency subspace according to the channel of the initial time slot at the beginning of each communication frame, generate the initial analog beamforming matrix, and configure it to the phase shifter network.

[0037] The simulated beam tracking module is used to update the main radio frequency subspace of the previous time slot in each time slot of the communication frame using incremental singular value decomposition, and generate the simulated beamforming matrix of the current time slot based on the updated main radio frequency subspace.

[0038] A digital beam inference module is used to generate a digital beamforming matrix for the current time slot based on the state evolution of a liquid neural network in each time slot within the communication frame.

[0039] A frame-level parameter correction module is used to correct the parameters of the liquid neural network at the end of each communication frame based on the accumulated loss function within that frame, so that the liquid neural network can adapt to channel changes.

[0040] In one embodiment, the frame-level parameter correction module is used to correct the trainable parameters of the liquid neural network at the end of each communication frame, based on the loss function accumulated across all time slots within the frame, using a backpropagation algorithm.

[0041] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-described spaceborne phased array hybrid digital-analog beamforming method through the computer program.

[0042] The technical solutions provided in this application embodiment may include the following beneficial effects:

[0043] This application, by reasonably dividing the time scale into frame level and time slot level, can accurately adapt to the dual dynamic changes of the channel's slow-changing characteristics and fast-changing instantaneous characteristics, achieving a good balance between computational complexity and beam tracking accuracy.

[0044] On the analog beamside, an incremental singular value decomposition mechanism is employed, which avoids the high computational overhead of performing full decomposition of the high-dimensional channel matrix for each time slot through subspace inheritance and incremental updates. On the digital beamside, a liquid neural network significantly reduces the frequency of model training and significantly improves computational efficiency by combining state evolution and parameter adjustment. The synergy of these two approaches enables a low-complexity, lightweight hybrid analog-digital beamforming design, significantly reducing the system's online operational burden while ensuring communication performance. Attached Figure Description

[0045] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0046] Figure 1 This is a flowchart of a spaceborne phased array hybrid digital-analog beamforming method according to an embodiment of this application;

[0047] Figure 2 This is a schematic diagram of a hybrid digital-analog beamforming architecture according to an embodiment of this application;

[0048] Figure 3 This is a schematic diagram of a dual-time-scale frame structure according to an embodiment of this application;

[0049] Figure 4 This is a schematic diagram of a simulated beamforming matrix update method according to an embodiment of this application;

[0050] Figure 5 This is a schematic diagram of the state evolution of a liquid neuron according to an embodiment of this application;

[0051] Figure 6 This is a schematic diagram of adjusting parameters of a liquid neural network according to an embodiment of this application;

[0052] Figure 7 This is a schematic diagram of an application scenario according to an embodiment of this application;

[0053] Figure 8 This is a schematic diagram of a spaceborne phased array hybrid digital-analog beamforming device according to an embodiment of this application;

[0054] Figure 9 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0055] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0056] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0057] The following is a detailed description of the spaceborne phased array hybrid digital-analog beamforming method according to embodiments of this application, with reference to the accompanying drawings. Figure 1 As shown, the method mainly includes the following steps:

[0058] The S101 is equipped with a satellite-borne phased array antenna transmission architecture and a dual-time-scale communication frame structure, which discretizes the communication time axis into frame-level and time-slot-level scales.

[0059] In one implementation, the transmission architecture of the spaceborne phased array antenna and the dual-time-scale communication frame structure are configured, including first initializing the downlink communication link configuration of the low-Earth orbit satellite and enabling the phased array antenna and the hybrid digital-analog beamforming architecture as the signal transmitter.

[0060] Specifically, consider a low-orbit satellite equipped with... Uniform planar array antenna with individual elements and Multiple radio frequency links simultaneously provide downlink services to multiple single-antenna ground users, among which and They are respectively and The number of antennas on each axis. The user set is denoted as... Satellites employ methods such as Figure 2 The hybrid analog-digital beamforming architecture shown consists of, from left to right, a data stream input module, a digital baseband processing module, multiple RF chains, and an analog beamforming network composed of phase shifters and signal synthesis units. The final signal is transmitted via a large-scale antenna array on the right. Following... Based on the hybrid analog-digital beamforming architecture, the satellite's transmitted signal at any transmission moment can be represented as:

[0061]

[0062] in, The simulated beamforming matrix is ​​implemented using a phase shifter network, satisfying the constant mode constraint; The digital beamforming matrix is ​​updated in the baseband domain. Let be the emission symbol vector, satisfying .

[0063] Furthermore, the time-domain parameters of the communication link are configured, and a dual-time-scale frame structure is constructed. The frame structure discretizes the communication time axis into a frame-level scale and a time-slot-level scale. The frame-level scale is used to characterize the slow-varying characteristics of the channel, and the time-slot-level scale is used to characterize the fast-fading characteristics of the channel.

[0064] Specifically, the high-speed motion of low-Earth orbit satellites causes drastic dual dynamic changes in the channel, such as... Figure 3 As shown, the communication timeline is divided into scales, including frame-level scale and slot-level scale. The frame-level scale considers... The time period of consecutive frames is denoted as . The duration of each frame is The time-slot level further divides each frame into... There are 1 time slot, and the duration of each time slot is 1. ,satisfy , No. The set of time slot indices within each frame is denoted as . Within each time slot, the instantaneous channel state is considered to remain unchanged.

[0065] Finally, based on the phased array antenna, the hybrid digital-analog beamforming architecture, and the dual-time-scale frame structure, the spectral efficiency optimization target of the system is calculated and established.

[0066] Specifically, in the The first frame The first time slot, for the first For each user, the received signal can be represented as:

[0067]

[0068] in, For this moment, the satellite is connected to the user. The channel response vector, The noise is additive white Gaussian noise. Channel response vector. It can be expressed as the superposition of direct and indirect radii, given by the following formula:

[0069]

[0070] in, Rice factor; For large-scale fading, mainly caused by satellites and users The free space loss between them is determined by the following formula:

[0071]

[0072] in, For wavelength, The speed of light in a vacuum For operating frequency, For satellites and users The theoretical distance between them. and These represent the direct trajectory component and the non-direct trajectory component, respectively. and These are the complex gains for the direct and indirect beams, respectively. It is a random phase; The planar matrix steering vector can be further expressed as:

[0073]

[0074] in, For Kronecker product; and respectively along shaft and The axial direction of the guide vector. and Let be the departure angle. At this point, the set of downlink channels for all users can be represented as: .

[0075] Based on the above signal model, the signal-to-interference-plus-noise ratio (SINR) of the received signal can be expressed as:

[0076]

[0077] Then the first The spectral efficiency for each user in this time slot is:

[0078]

[0079] This application, by reasonably dividing the dual time scales, can accurately adapt to the dual dynamic changes of the channel's slow-changing characteristics and fast-changing instantaneous characteristics, achieving a good balance between computational complexity and beam tracking accuracy.

[0080] S102 extracts the main radio frequency subspace based on the channel of the initial time slot at the beginning of each communication frame, generates an initial analog beamforming matrix, and configures it to the phase shifter network.

[0081] In one implementation, at the beginning stage of each communication frame, the main radio frequency subspace is extracted based on the channel of the initial time slot, an initial simulated beamforming matrix is ​​generated and configured for the phase shifter network, including firstly, at the beginning stage of each communication frame, obtaining the instantaneous channel matrix, performing singular value decomposition on the channel matrix, and extracting the basis matrix of the main radio frequency subspace.

[0082] Specifically, singular value decomposition is performed on the instantaneous channel matrix. The result of the singular value decomposition can be given by the following formula: .

[0083] in, It is a left singular vector matrix. Let be a diagonal matrix consisting of singular values ​​arranged in descending order. It is a right singular vector matrix.

[0084] Extracting the front of the left singular vector matrix The columns form the basis matrix of the principal radio frequency subspace to maximize the retention of channel energy. Specifically, the principal subspace basis matrix can be represented as:

[0085] ;

[0086] Its column vectors correspond to the channels with the largest energy. There are orthogonal directions, and the corresponding singular value matrix is: .

[0087] Furthermore, an optimization problem is established based on the subspace approximation criterion, and the optimization problem is solved under constant mode constraints to obtain the initial simulated beamforming matrix.

[0088] Specifically, considering that in a mixed-signal architecture, analog beamforming is implemented by a phase shifter network, the amplitude of its elements is limited by constant mode constraints. To make the actually generated beam approximate the ideal RF subspace as closely as possible, a subspace approximation-based design criterion is adopted, transforming the construction problem of the analog beamforming matrix into the following optimization problem:

[0089]

[0090] in, Let represent the Frobenius norm. Under constant modulus constraints, its optimal solution is . The simulated beamforming matrix can then be obtained from the following equation: .

[0091] Finally, corresponding phase control commands are generated based on the initial simulated beamforming matrix and sent to the phase shifter network.

[0092] According to the calculation The corresponding phase control commands are generated and sent to the phase shifter network to complete the initial analog beam configuration.

[0093] S103 updates the main radio frequency subspace of the previous time slot in each time slot within the communication frame using incremental singular value decomposition, and generates the analog beamforming matrix of the current time slot based on the updated main radio frequency subspace.

[0094] In one implementation, when a new channel sample arrives, the newly added channel features are extracted based on the main radio frequency subspace of the previous time slot, and a first increment matrix is ​​constructed. This includes calculating the channel projection and orthogonal residual components of the current time slot based on the main radio frequency subspace of the previous time slot when a new channel sample arrives.

[0095] Specifically, consider the first The first frame. Assuming the first frame... Subspace estimation has been obtained for each time slot. and the corresponding singular value matrix When new channel samples Upon arrival, update the simulated beamforming matrix, including the first beam within the frame. For each time slot, the channel projection and orthogonal residual components of the current time slot are calculated based on the main radio frequency subspace of the previous time slot.

[0096] Specifically, Projection matrix in the old subspace The residual matrix in the complement space of the old subspace. They can be represented as:

[0097]

[0098]

[0099] The residual matrix It includes new channel features that were not covered by the old subspace.

[0100] Furthermore, the channel projection and orthogonal residual components are orthogonally normalized to obtain the new channel features.

[0101] Specifically, the orthogonally normalized residual basis matrix and residual intensity matrix The results can be calculated using the following formulas:

[0102]

[0103]

[0104] Furthermore, based on the newly added channel features and by introducing a forgetting factor, combined with historical singular value information, a first incremental matrix is ​​constructed.

[0105] Specifically, in order to balance the weights of historical information and current instantaneous information during the update process and achieve smooth tracking of time-varying channels, a forgetting factor is introduced. ,satisfy And construct a low-dimensional first increment matrix. The first increment matrix can be given by the following equation:

[0106]

[0107] Furthermore, singular value decomposition is performed on the first incremental matrix, and the basis matrix of the main radio frequency subspace of the previous time slot is updated.

[0108] Specifically, in order to obtain the updated optimal subspace, the first increment matrix is... Perform singular value decomposition, i.e. The subspace is then updated based on the eigenvalue decomposition results. The updated subspace can be given by the following formula:

[0109]

[0110] in, Indicates taking The former Column, corresponding to the largest There are 1 singular value. The corresponding 1 / 2 is retained. The largest singular values ​​form a diagonal matrix. .

[0111] Finally, based on the updated main radio frequency subspace basis matrix, the simulated beamforming matrix for the current time slot is constructed under constant mode constraints.

[0112] Specifically, to adapt to the hardware limitations of the phase shifter network, analog beamforming design needs to be performed while satisfying the constant mode constraint. Therefore, the first... The first frame, the first The simulated beamforming matrix for each time slot can be given by the following formula:

[0113]

[0114] To facilitate understanding of the simulated beam update method in this application, the following is in conjunction with the appendix. Figure 4 Further description. For example... Figure 4 As shown, the process includes: when a new channel sample arrives, calculating the channel projection and orthogonal residual components of the current time slot based on the main radio frequency subspace of the previous time slot; performing orthogonal normalization on the projection residuals to extract new channel features independent of the old subspace; introducing a forgetting factor and combining it with historical singular value information to construct an incremental matrix; performing singular value decomposition on the incremental matrix and updating the main radio frequency subspace basis matrix; and constructing the simulated beamforming matrix of the current time slot under constant mode constraints based on the online updated main radio frequency subspace basis matrix.

[0115] Through this series of operations, it is possible to quickly track channel changes and adjust the analog beam pointing in real time without repeatedly performing singular value decomposition of the high-dimensional channel matrix, ensuring that the beam is always pointed at the user.

[0116] S104 generates the digital beamforming matrix for the current time slot based on the state evolution of a liquid neural network in each time slot within the communication frame.

[0117] In one implementation, in each time slot within a communication frame, a digital beamforming matrix for the current time slot is generated based on a liquid neural network. This includes first initializing the current state of the liquid neural network using a hot-start mechanism, and then calculating the gradient of the instantaneous loss function based on the beamforming optimization variables of the previous time slot.

[0118] Specifically, proceed to the... Each frame Each time slot employs a warm-start strategy, directly inheriting the digital beamforming basis matrix from the end of the previous time slot. Digital beamforming matrix and network hidden state This serves as the current initial point. At this point, the loss function can be calculated using the following formula:

[0119]

[0120] In addition, considering the limited satellite launch power, a power penalty term is introduced. This represents the coefficient of the power penalty term. The gradient of the loss function is then calculated. Input liquid neural network.

[0121] Furthermore, the gradient of the instantaneous loss function is input into the liquid neural network to drive the liquid neural network to perform continuous-time hidden state evolution.

[0122] This liquid neural network consists of three layers of liquid neurons: an input layer, a command layer, and an action layer. The output of the previous layer serves as the input to the next layer, and the input to the input layer comes from external input. For example... Figure 5 As shown, the hidden state of each liquid neuron is updated according to the following formula:

[0123]

[0124] in, and These represent the hidden state and input of the neuron, respectively. , and Each means respectively by Nonlinear mapping function for control; It is a set of trainable parameters, which are kept constant within the frame to reduce training overhead; It is an element-wise sigmoid activation function; For Hadama accumulation.

[0125] Furthermore, based on the evolved hidden state, a second incremental matrix is ​​output to update the digital beam.

[0126] After the nonlinear evolution of the liquid neuron, the output layer of the neural network maps the current hidden state to a second increment matrix. Here This represents the optimal update step size determined by the network based on the current gradient direction and historical information.

[0127] Finally, the digital beamforming basis matrix is ​​updated based on the second incremental matrix to generate the digital beamforming matrix for the current time slot.

[0128] Specifically, the second increment matrix output by the network is used for updating to obtain the basis matrix of the current time slot, i.e. Subsequently, using the updated Obtain the original digital beamforming matrix ,in It is a low-dimensional equivalent baseband channel.

[0129] The generated digital beamforming matrix undergoes power normalization to meet the satellite's maximum transmit power constraint. Since the generated digital beamforming matrix may not meet the hardware power limits, normalization is necessary. When the satellite's maximum total transmit power is... At that time, the final digital beamforming matrix It can be represented as:

[0130]

[0131] At the end of each communication frame, S105 modifies the parameters of the liquid neural network based on the accumulated loss function within that frame, so that the liquid neural network can adapt to channel changes.

[0132] At the end of each communication frame, the trainable parameters of the liquid neural network are corrected using the backpropagation algorithm based on the accumulated loss function of all time slots within that frame.

[0133] Specifically, such as Figure 6 As shown, in order for the liquid neural network to continuously learn and adapt to changes in the satellite channel, this step utilizes frame-level cumulative error to adjust the network parameters. Update. In the... During the transmission of a communication frame, for each time slot within that frame, the updated digital beamforming matrix is ​​directly used as the basis for the transmission. Calculate the instantaneous loss function, which reflects the actual performance of the current beamforming strategy. After all time slots of the frame have been processed, calculate the frame-level cumulative loss function. Then, backpropagation is performed, and the parameters are updated using the adaptive moment estimation optimization criterion. The specific update formula is as follows:

[0134]

[0135] in, This is the learning rate.

[0136] The spaceborne phased array hybrid digital-analog beamforming method proposed in this application has applications such as... Figure 7 As shown, the large gray planar array antenna features densely packed array elements representing a large-scale phased array structure. Multiple beams radiate downwards from the antenna array, precisely pointing towards ground user terminal equipment within multiple dashed circles, vividly demonstrating the multi-beam spatial multiplexing and directional coverage capabilities achieved through hybrid analog-digital beamforming technology. The overall design clearly conveys the core working mechanism of establishing dynamic, multi-target communication links between the satellite and ground users via the phased array antenna.

[0137] This application proposes a hybrid digital-analog beamforming design based on dual time scales, which better meets the needs of multi-user services and improves spectrum efficiency. The advantage of this method lies in its ability to accurately adapt to the dual dynamic changes of the channel through dual time scale design, achieving a good balance between computational overhead and beam tracking accuracy. Simultaneously, it avoids full high-dimensional matrix decomposition through subspace inheritance and incremental update mechanisms, and significantly reduces the frequency of model training by combining liquid neural network state evolution and parameter adjustment. These two approaches work together to achieve a low-complexity, lightweight hybrid digital-analog beamforming design, significantly reducing the online operating burden of the system while ensuring communication performance.

[0138] According to another aspect of the embodiments of this application, a spaceborne phased array hybrid beamforming apparatus for implementing the above-described spaceborne phased array digital-analog hybrid beamforming method is also provided. For example... Figure 8 As shown, the device includes:

[0139] The communication link configuration module 801 is used to configure the transmission architecture of the spaceborne phased array antenna and the dual-time-scale communication frame structure. The frame structure discretizes the communication time axis into frame-level scale and time-slot-level scale.

[0140] The analog beam initialization module 802 is used to extract the main radio frequency subspace according to the channel of the initial time slot at the beginning stage of each communication frame, generate an initial analog beamforming matrix and configure it to the phase shifter network.

[0141] The analog beam tracking module 803 is used to update the main radio frequency subspace of the previous time slot in each time slot within the communication frame using incremental singular value decomposition, and generate the analog beamforming matrix of the current time slot based on the updated main radio frequency subspace.

[0142] The digital beam inference module 804 is used to generate the digital beamforming matrix of the current time slot based on the state evolution of the liquid neural network in each time slot within the communication frame.

[0143] The frame-level parameter correction module 805 is used to correct the parameters of the liquid neural network at the end of each communication frame based on the accumulated loss function within that frame, so that the liquid neural network can adapt to channel changes.

[0144] In one implementation, a frame-level parameter correction module is used to correct the trainable parameters of the liquid neural network at the end of each communication frame, based on the loss function accumulated across all time slots within that frame, using a backpropagation algorithm.

[0145] It should be noted that the spaceborne phased array hybrid beamforming device provided in the above embodiments is only illustrated by the division of the above functional modules when executing the spaceborne phased array hybrid beamforming method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the spaceborne phased array hybrid beamforming device and the spaceborne phased array hybrid beamforming method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0146] According to another aspect of the embodiments of this application, an electronic device corresponding to the spaceborne phased array mixed digital-analog beamforming method provided in the foregoing embodiments is also provided, so as to execute the above-described spaceborne phased array mixed digital-analog beamforming method.

[0147] Please refer to Figure 9 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 9 As shown, the electronic device includes: a processor 900, a memory 901, a bus 902, and a communication interface 903. The processor 900, the communication interface 903, and the memory 901 are connected via the bus 902. The memory 901 stores a computer program that can run on the processor 900. When the processor 900 runs the computer program, it executes the spaceborne phased array hybrid digital-analog beamforming method provided in any of the foregoing embodiments of this application.

[0148] The memory 901 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 903 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0149] Bus 902 can be an ISA bus, PCI bus, or EISA bus, etc. Buses can be divided into address buses, data buses, control buses, etc. Memory 901 is used to store programs. After receiving execution instructions, processor 900 executes the programs. The spaceborne phased array hybrid digital-analog beamforming method disclosed in any of the aforementioned embodiments of this application can be applied to processor 900, or implemented by processor 900.

[0150] The processor 900 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 900 or by instructions in software form. The processor 900 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 901. Processor 900 reads the information in memory 901 and, in conjunction with its hardware, completes the steps of the above method.

[0151] The electronic device provided in this application embodiment and the spaceborne phased array hybrid digital-analog beamforming method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0152] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0153] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A spaceborne phased array hybrid digital-analog beamforming method, characterized in that, include: The system is configured with a satellite-borne phased array antenna transmission architecture and a dual-time-scale communication frame structure. The frame structure discretizes the communication time axis into a frame-level scale and a time-slot-level scale. The high-speed motion of the low-orbit satellite causes the channel to exhibit dual dynamic changes. The frame-level scale is used to characterize the slow-varying characteristics of the channel, and the time-slot-level scale is used to characterize the fast fading characteristics of the channel. At the beginning of each communication frame, the main radio frequency subspace is extracted according to the channel of the initial time slot, an initial simulated beamforming matrix is ​​generated and configured for the phase shifter network; In each time slot within the communication frame, incremental singular value decomposition is used to update the main radio frequency subspace of the previous time slot, and the simulated beamforming matrix of the current time slot is generated based on the updated main radio frequency subspace. In each time slot within the communication frame, the digital beamforming matrix for the current time slot is generated based on the state evolution of the liquid neural network. This includes: initializing the current state of the liquid neural network using a hot-start mechanism and calculating the instantaneous loss function gradient based on the beamforming optimization variables of the previous time slot; inputting the instantaneous loss function gradient into the liquid neural network to drive the liquid neural network to perform continuous-time hidden state evolution; outputting a second increment matrix for updating the digital beam according to the evolved hidden state; and updating the digital beamforming basis matrix based on the second increment matrix to generate the digital beamforming matrix for the current time slot. At the end of each communication frame, the parameters of the liquid neural network are modified based on the accumulated loss function within that frame to enable the liquid neural network to adapt to channel changes. This includes: at the end of each communication frame, the trainable parameters of the liquid neural network are modified using a backpropagation algorithm based on the accumulated loss function across all time slots within that frame.

2. The method according to claim 1, characterized in that, The transmission architecture configured with a spaceborne phased array antenna and a dual-time-scale communication frame structure includes: Initialize the downlink communication link configuration of the low-Earth orbit satellite and enable the phased array antenna and hybrid digital-analog beamforming architecture as the signal transmitter; Configure the time-domain parameters of the communication link and construct a dual-time-scale frame structure, wherein the frame structure discretizes the communication time axis into frame-level scale and time-slot-level scale; Based on the phased array antenna, the hybrid digital-analog beamforming architecture, and the dual-time-scale frame structure, the spectral efficiency optimization target of the system is calculated and established.

3. The method according to claim 1, characterized in that, At the beginning of each communication frame, the main radio frequency subspace is extracted based on the channel of the initial time slot, an initial analog beamforming matrix is ​​generated and configured for the phase shifter network, including: At the beginning of each communication frame, the instantaneous channel matrix is ​​acquired, and singular value decomposition is performed on the channel matrix to extract the basis matrix of the main radio frequency subspace. An optimization problem is established based on the subspace approximation criterion, and the optimization problem is solved under constant mode constraints to obtain the initial simulated beamforming matrix. The corresponding phase control commands are generated based on the initial simulated beamforming matrix and sent to the phase shifter network.

4. The method according to claim 1, characterized in that, In each time slot within the communication frame, incremental singular value decomposition is used to update the main radio frequency subspace of the previous time slot. Based on the updated main radio frequency subspace, the analog beamforming matrix of the current time slot is generated, including: When a new channel sample arrives, the new channel features are extracted based on the main radio frequency subspace of the previous time slot, and the first incremental matrix is ​​constructed. Perform singular value decomposition on the first incremental matrix and update the main radio frequency subspace basis matrix of the previous time slot; Based on the updated main radio frequency subspace basis matrix, the simulated beamforming matrix for the current time slot is constructed under constant mode constraints.

5. The method according to claim 4, characterized in that, When a new channel sample arrives, based on the main radio frequency subspace of the previous time slot, the newly added channel features are extracted, and the first incremental matrix is ​​constructed, including: When a new channel sample arrives, the channel projection and orthogonal residual components of the current time slot are calculated based on the main radio frequency subspace of the previous time slot. The channel projection and orthogonal residual components are orthogonally normalized to obtain the new channel features; Based on the newly added channel features, and by introducing a forgetting factor and combining historical singular value information, the first incremental matrix is ​​constructed.

6. A spaceborne phased array hybrid digital-analog beamforming device, characterized in that, include: The communication link configuration module is used to configure the transmission architecture of the spaceborne phased array antenna and the dual-time-scale communication frame structure. The frame structure discretizes the communication time axis into frame-level scale and time-slot-level scale. The high-speed motion of the low-orbit satellite causes the channel to exhibit dual dynamic changes. The frame-level scale is used to characterize the slow-varying characteristics of the channel, and the time-slot-level scale is used to characterize the fast fading characteristics of the channel. The analog beam initialization module is used to extract the main radio frequency subspace according to the channel of the initial time slot at the beginning of each communication frame, generate the initial analog beamforming matrix, and configure it to the phase shifter network. The simulated beam tracking module is used to update the main radio frequency subspace of the previous time slot in each time slot of the communication frame using incremental singular value decomposition, and generate the simulated beamforming matrix of the current time slot based on the updated main radio frequency subspace. A digital beam inference module is used to generate a digital beamforming matrix for the current time slot based on the state evolution of a liquid neural network in each time slot within the communication frame. This includes: initializing the current state of the liquid neural network using a hot-start mechanism and calculating the gradient of the instantaneous loss function based on the beamforming optimization variables of the previous time slot; inputting the instantaneous loss function gradient into the liquid neural network to drive continuous-time hidden state evolution; outputting a second increment matrix for updating the digital beam according to the evolved hidden state; and updating the digital beamforming basis matrix based on the second increment matrix to generate the digital beamforming matrix for the current time slot. A frame-level parameter correction module is used to correct the parameters of the liquid neural network at the end of each communication frame based on the accumulated loss function within that frame, so that the liquid neural network can adapt to channel changes. This includes: at the end of each communication frame, correcting the trainable parameters of the liquid neural network using a backpropagation algorithm based on the accumulated loss function of all time slots within that frame.

7. An electronic device, characterized in that, It includes a processor and a memory storing program instructions, the processor being configured to perform the spaceborne phased array hybrid digital-analog beamforming method as described in any one of claims 1 to 5 when executing the program instructions.

Images