Predictive cooperative beam forming and resource allocation method based on cellular-free multi-station information fusion
By employing a predictive cooperative beamforming and resource allocation method based on multi-station information fusion in a non-cellular system, the problem of high beam tracking overhead in a large-scale MIMO UAV communication system is solved, achieving low training overhead and high system throughput improvement, and is applicable to both TDD and FDD systems.
Patent Information
- Application Number
- CN202511783966.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-01
- Publication Date
- 2026-02-13
AI Technical Summary
Beam tracking overhead is too high in non-cellular massive MIMO UAV communication systems, and there is insufficient research on improving beam tracking performance through collaborative sensing capabilities in non-cellular scenarios, especially in TDD and FDD systems.
A predictive cooperative beamforming and resource allocation method based on non-cellular multi-station information fusion is adopted. By establishing uplink beam training and downlink data transmission models, EKF is used for state estimation and covariance information update, and information fusion is performed by CPU. The predictive posterior Cramer-Rao bound of tracking error is derived, and a resource allocation problem is constructed to optimize pilot length, user association and power allocation strategies, so as to achieve predictive beamforming.
It effectively reduces beam training overhead, increases system throughput, and is applicable to both TDD and FDD systems, improving estimation accuracy and communication efficiency.
Smart Images
Figure CN121531385A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of wireless communication and unmanned aerial vehicle network, and particularly relates to a prediction type cooperative beamforming and resource allocation method based on non-cell multi-station information fusion. BACKGROUND
[0002] With the continuous upgrading of ultra-low latency communication, global coverage and ultra-high density connection demand, the 6th Generation (6G) mobile communication technology emerges as the times require. As one of the key core technologies of 6G, non-cell large-scale MIMO (Multiple-Input Multiple-Out-put, MIMO) technology connects distributed base stations to the central processing unit (CPU) through the front link to obtain cooperative gain, thereby providing more consistent high-speed communication services. In recent years, with the vigorous development of low-altitude economy, unmanned aerial vehicle communication as an important component of space-ground integrated network has also attracted widespread attention due to its advantages such as line-of-sight propagation and flexible deployment. Therefore, considering the unmanned aerial vehicle communication based on non-cell large-scale MIMO has great application prospect and development potential.
[0003] However, in order to guarantee the high-speed air communication demand, the distributed base station with large-scale antenna needs to obtain channel state information through high-overhead beam training, and the high mobility of unmanned aerial vehicles leads to a significant shortening of channel coherence time, which will further increase the training overhead required by the system. Some scholars have proposed to use location information to assist in reducing the training overhead for high-mobility unmanned aerial vehicle communication scenarios. Specifically, by estimating the parameters of the received communication echo signal at the base station and tracking the mobile user based on the extended Kalman filter (EKF), a predictive beamforming is designed based on the tracked location information, thereby avoiding frequent beam training overhead. However, this scheme is based on the need for full-duplex physical structure of the communication system for echo signals, and the interference between echo signals and communication signals will face challenges; and most of the research focuses on point-to-point or point-to-multipoint beam tracking, and the research on improving the beam tracking performance using cooperative sensing capability in the non-cell scenario is still insufficient. SUMMARY
[0004] The application aims to solve the problem of high beam tracking overhead in a cell-free massive MIMO unmanned aerial vehicle communication system, and provides a cooperative beam tracking method with low training overhead and low backhaul overhead. Compared with the traditional method, the method achieves a good compromise between reducing training overhead and improving system throughput, and can be applied to time division duplexing (TDD) and frequency division duplexing (FDD) systems at the same time by utilizing spatial uplink-downlink reciprocity.
[0005] The application provides a prediction type cooperative beam forming and resource allocation method based on cell-free multi-station information fusion, comprising the following steps: Step one: establish an uplink beam training and downlink data transmission model of a cell-free massive MIMO assisted unmanned aerial vehicle communication system, and construct a downlink effective and spectral efficiency model.
[0006] Step two: using the proposed beam tracking frame structure, the base station side uses the uplink pilot received signal as the measurement based on EKF to realize the update of the unmanned aerial vehicle user state estimation and covariance information.
[0007] Step three: each base station transmits the updated state estimation and covariance information to the CPU through the backhaul link for information fusion, and derives the prediction posterior Cramér-Rao bound (PCRB) of the tracking error and the explicit analytical expression of the uplink pilot length.
[0008] Step four: based on the analytical expression obtained in step three, the CPU constructs a resource allocation problem with the maximum effective and spectral efficiency as the target, with the estimation error precision requirement and the power-correlation limit as the constraint conditions, selects a suitable combinatorial optimization algorithm, and solves to obtain the optimal uplink pilot length, downlink base station user association scheme and power allocation strategy.
[0009] Step five: the CPU sends the fused state estimation and covariance in step three and the resource allocation result in step four to each base station, and the base station side performs state estimation and covariance prediction, and performs predictive beam forming to complete downlink data transmission.
[0010] Further, in step one, an uplink beam training and downlink data transmission model of a cell-free massive MIMO assisted unmanned aerial vehicle communication system is established, and a downlink effective and spectral efficiency model is constructed.
[0011] It is assumed that in the coverage area of interest, there are ground base stations serving unmanned aerial vehicle users, and the ground base stations are each equipped with an array of a uniform planar array, where with respectively the number of horizontal and vertical antennas, single antenna at UAV users; ground base stations are connected to CPU for centralized resource scheduling through wireless backhaul links; at time slot , UAV users send length orthogonal pilot signals to AP, the th base station uplink received signal is: ; where is the continuous time variable of uplink received signal, is the uplink transmit power of the th UAV user, is the channel between the th base station and the th UAV user, dominated by line-of-sight, transmit pulse shaping signal, is the delay between the th base station and the th UAV user, is the base station side Gaussian white noise; by projecting the uplink received signal onto different orthogonal pilots, the received signal of different UAV users is: ; where is the received noise after projection; Then consider the downlink data transmission phase, at time slot , the downlink received signal of the th UAV user is: ; where , and respectively represent the th base station associated variable of the th UAV user, downlink transmit power and transmit beamforming vector, by the predictive beamforming construction in step five, is the downlink transmission signal with unit power, is the user side Gaussian white noise; thus the effective and spectral efficiency of the communication system is: ; where and Each time slot and symbol time are represented separately. This represents the noise power on the user side.
[0012] Furthermore, step two employs the proposed beam tracking frame structure, such as... Figure 2 As shown, unlike the traditional frame structure, the proposed beam tracking frame structure contains each frame... In each time slot, the drone user only needs to perform uplink beam training in the first time slot. The base station uses the uplink pilot signal received by EKF as a measurement to update the drone user's state estimation and covariance information.
[0013] Let the first The state vector of each drone user is: ,in and For drone users in time slots The position and velocity components are obtained by parameter estimation of the projected uplink pilot received signal using the maximum likelihood method. The base station regarding the first Measurement vectors of drone users Due to the nonlinearity between the measurement vector and the state vector, and because the measurement information of each UAV user can be distinguished by orthogonal pilots, distributed beam tracking can be performed using EKF at each base station side; with the first... Taking drone user status tracking as an example, the core steps of extended Kalman filtering include two stages: prediction and update. The prediction process includes state estimation and covariance prediction: ; ; in and These are the state estimate and covariance matrix after information fusion, respectively. and These are the predicted state estimate and the covariance matrix, respectively. The motion state transition equation for the UAV is as follows: The state noise covariance; The update process includes Kalman gain calculation and state estimation covariance update: ; ; ; in For Kalman gain, For the measurement equation, To measure the linearization matrix of the equation, To measure the noise covariance, and The first The measurement at the first base station obtained the first Individual drone user state estimation and covariance matrix update It is an identity matrix.
[0014] Furthermore, in step three, each base station transmits the updated state estimate and covariance information to the CPU via the backhaul link for information fusion, and derives explicit analytical expressions for the predictive posterior Cramer-Rao bound (PCRB) of the tracking error and the uplink pilot length: Set the first The state estimates and covariances of each drone user are: and The CPU uses the covariance cross criterion for information fusion. ; ; in The fusion coefficients reflect the confidence level of the corresponding local posterior estimate information and satisfy the following: ; Then, the predictive PCRB considering the information fusion process is derived. for: ; in Let be the Fisher information matrix of the previous time slot, and be the reciprocal of PCRB. Defined based on the relationship between the measurement noise covariance and the uplink pilot length. , ,in for: ; in The noise proportionality coefficients were measured for time delay, Doppler, azimuth, and elevation angle parameters, respectively; then... Eigenvalue decomposition yields: ; in The eigenvector matrix, The matrix is a diagonal matrix composed of eigenvalues; the derivation yields: ; in for The One element, for diagonal Each element.
[0015] Further, step four is to build a resource allocation problem based on the analytical expression obtained in step three, with the objective of maximizing the effective and spectral efficiency, and the constraints of the estimation error accuracy requirement and the power-related limitations. A suitable combinatorial optimization algorithm is selected to solve the problem and obtain the optimal uplink pilot length, downlink base station-user association scheme, and power allocation strategy.
[0016] The resource allocation problem built at the CPU is as follows: ; where , represents the set of association variables and power allocation variables, is the tracking error metric defined according to the explicit analytical expression between the predicted PCRB and the uplink pilot length obtained in step three: ; is the tracking error requirement, is a set of positive integers, and are the maximum number of users served by each base station and the maximum transmit power, respectively; Since the tracking error constraint in this resource allocation optimization problem is only related to the pilot length, and the pilot length has a monotonic relationship with the effective and spectral efficiency, the optimal uplink pilot length can be obtained by solving the tracking error metric expression first. After fixing the optimal pilot length, the remaining problem is a mixed integer non-convex optimization problem, which is solved iteratively with low complexity using compressed sensing and fractional programming. By introducing compressed sensing relaxation association scalar: ; where denotes the 0-norm, i.e., the number of non-zero values, is a weighting factor, which is updated using the following iterative method: ; where denotes the number of iterations, is a smoothing factor that can ensure numerical stability at each iteration. The core step of fractional programming used in each iteration process is: ; ; ; where , and These are, respectively, the signal-to-interference-plus-noise ratio (SINR) auxiliary variable, the quadratic transformation auxiliary variable in fractional programming, and the power update variable. For the effective and spectral efficiency coefficients, where The optimal uplink pilot length is obtained by solving the expression for the tracking error metric. and These are the Lagrange multipliers corresponding to the user association constraints and power constraints of the base station; Therefore, the optimal uplink pilot length is first calculated using the tracking error metric expression. Then update sequentially using an iterative method. , , The optimal resource allocation result is obtained when the objective function converges.
[0017] Furthermore, in step five, the CPU sends the fused state estimate and covariance from step three, as well as the resource allocation results from step four, to each base station. The base station performs state estimation and covariance prediction, and performs predictive beamforming to complete downlink data transmission.
[0018] No. After obtaining the fusion information and resource allocation results, the first base station first predicts the th base station based on the state equation. Status prediction for individual drone users Then, the channel is reconstructed using a line-of-sight-dominated channel model based on the location of the ground base station. And thus obtain the first Reconstructed channel matrix for all users at each base station To reduce inter-user interference, zero-forcing precoding is used to achieve predictive transmit beamforming. : ; in No. The base station for the first The transmit beamforming vector of each UAV; then the base station side based on Perform downlink data transmission until the next beam tracking frame, then repeat steps two through five.
[0019] This invention proposes a predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion. Compared with traditional methods, the proposed beam tracking frame structure can effectively reduce beam training overhead, and the low backhaul overhead information fusion method improves estimation accuracy, thereby further increasing system throughput. In addition, this invention utilizes the uplink and downlink spatial reciprocity of the channel, and can be applied to both TDD and FDD systems. Attached Figure Description
[0020] Figure 1A flowchart illustrating a predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion, provided for an embodiment of the present invention; Figure 2 This is a schematic diagram of a low-overhead beam tracking frame structure provided in an embodiment of the present invention, wherein... Figure 2 (a) in the image represents the traditional beam tracking frame structure. Figure 2 (b) in the figure represents the proposed low-overhead beam tracking frame structure; Figure 3 This invention provides a scenario setting and trajectory tracking result diagram for a non-cellular MIMO unmanned aerial vehicle communication system according to an embodiment of the present invention. Figure 4 A comparison of the tracking error performance of a predictive cooperative beamforming and resource allocation method based on non-cellular multi-station information fusion provided in this embodiment of the invention with that of a traditional frame structure that does not include information fusion method; Figure 5 The diagram shows a comparison of the effectiveness and spectral efficiency of a predictive cooperative beamforming and resource allocation method based on non-cellular multi-station information fusion provided in this embodiment of the invention, and a traditional frame structure without information fusion based on a uniform resource allocation method. Detailed Implementation
[0021] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0022] The following is an embodiment of the present invention suitable for a predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion: Step 1: Establish uplink beam training and downlink data transmission models for a non-cellular massive MIMO-assisted UAV communication system, and construct downlink efficiency and spectral efficiency models.
[0023] Step 2: Using the proposed beam tracking frame structure, the base station uses the uplink pilot signal received by EKF as a measurement to update the UAV user state estimation and covariance information.
[0024] Step 3: Each base station transmits the updated state estimate and covariance information to the CPU via the backhaul link for information fusion, and derives explicit analytical expressions for the predictive posterior Cramer-Rao bound (PCRB) of the tracking error and the uplink pilot length.
[0025] Step 4: Based on the analytical expression obtained in Step 3, the CPU constructs a resource allocation problem with the goal of maximizing efficiency and spectral efficiency, and with the constraints of estimation error accuracy requirements and power-association limitations. A suitable combinatorial optimization algorithm is selected to solve the problem and obtain the optimal uplink pilot length, downlink base station user association scheme, and power allocation strategy.
[0026] Step 5: The CPU sends the fused state estimate and covariance from Step 3, as well as the resource allocation results from Step 4, to each base station. The base station performs state estimation and covariance prediction, and performs predictive beamforming to complete downlink data transmission.
[0027] In step one, an uplink beam training and downlink data transmission model for a non-cellular massive MIMO-assisted UAV communication system is established, and a downlink efficiency and spectral efficiency model is constructed.
[0028] Assuming that within the area of interest, there are One ground base station The service is provided to each drone user, and the number of array elements deployed at each ground base station is as follows: A uniform planar array, wherein and These represent the number of antennas in the horizontal and vertical directions, respectively, with each drone user having a single antenna; the ground base station is connected to the CPU via a wireless backhaul link for centralized resource scheduling.
[0029] Due to the sparsity of millimeter waves and the fact that ground-to-air scenarios typically lack obstructions, it is assumed that the channel is line-of-sight dominated, and in time slots... , No. The base station and the first The channel between individual drone users can be modeled as follows: ; in For path gain, For the path loss at the reference distance, , , , and The first The base station and the first Distance between drone users, Doppler shift, time delay, departure azimuth and departure-arrival angles. Represents the array manifold vector: ; ; and These are the array manifold vectors for the horizontal and pitch directions, respectively.
[0030] In the time slot The length sent by the drone user is Orthogonal pilot signal Give AP, number Uplink received signal from each base station for: ; in For the continuous time variable of the uplink received signal, For the first Uplink transmission power of each drone user For the first The base station and the first The communication channels between individual drone users are dominated by line-of-sight. Transmit pulse shaping signal, For the first The base station and the first latency between drone users The base station side uses Gaussian white noise; by projecting the uplink received signal onto different orthogonal pilots, different drone user received signals are obtained. for: ; in This refers to the received noise after projection. Next, consider the downlink data transmission phase, in the time slot. , No. Downlink signal received at a drone user's location for: ; in , and Representing the first The base station for the first The variables associated with each UAV include downlink transmit power and transmit beamforming vector. Through the predictive beamforming construction in step five, For unit power downlink transmission signals, The user-side Gaussian white noise is used to obtain the effective and spectral efficiency of the communication system. for: ; in and Each time slot and symbol time are represented separately. This represents the noise power on the user side.
[0031] In this embodiment of the invention, the number of base stations in a scenario of a non-cellular massive MIMO-assisted unmanned aerial vehicle communication system is set. Each base station is configured with the following number of antennas: Number of drone users Unit path loss coefficient The noise power spectral density is -174 dBm / Hz.
[0032] Step two employs the proposed beam tracking frame structure, such as... Figure 2 As shown, unlike the traditional frame structure, the proposed beam tracking frame structure contains each frame... In each time slot, the drone user only needs to perform uplink beam training in the first time slot. The base station uses the uplink pilot signal received by EKF as a measurement to update the drone user's state estimation and covariance information.
[0033] Assume the location set of ground base stations Fixed and known, the first The state vector of each drone user is: ,in and For drone users in time slots Regarding the position and velocity components, in this embodiment of the invention, it is assumed that the drone user is moving at a constant velocity in a straight line, and the system state equation is: ; in Here is the state transition matrix. The noise is state noise, which follows a zero-mean Gaussian distribution with a covariance of , and These are the noise equations for position and velocity, respectively.
[0034] The parameters are estimated by using the maximum likelihood method on the projected uplink pilot received signal to obtain the first... The base station regarding the first Measurement vectors of drone users Its relationship with the state vector is as follows: ; Therefore, the following measurement equation is obtained: ; in The nonlinear measurement equation represents the relationship between the measurement vector and the state vector. To measure noise, it follows a zero-mean Gaussian distribution with a covariance of ,in Let represent the measurement noise of time delay, Doppler, azimuth, and elevation angles, respectively. We can assume that this noise is proportional to the signal-to-noise ratio of the projected uplink pilot received signal, i.e.: ; in This is the corresponding scaling factor, which is related to system configuration, transmission waveform, and other settings.
[0035] Because of the nonlinearity between the measurement vector and the state vector, and because the measurement information of each UAV user can be distinguished by orthogonal pilots, distributed beam tracking can be performed using EKF at each base station side; Taking drone user status tracking as an example, the core steps of extended Kalman filtering include two stages: prediction and update. The prediction process includes state estimation and covariance prediction: ; ; in and These are the state estimate and covariance matrix after information fusion, respectively. and These are the predicted state estimate and the covariance matrix, respectively. The motion state transition equation for the UAV is as follows: The state noise covariance; The update process includes Kalman gain calculation and state estimation covariance update: ; ; ; in For Kalman gain, For the measurement equation, To measure the linearization matrix of the equation, To measure the noise covariance, and The first The measurement at the first base station obtained the first Individual drone user state estimation and covariance matrix update It is an identity matrix.
[0036] In this embodiment of the invention, the length of each time slot is set. Each frame contains The observation time is set to 30 seconds, with the base station positions set at (0, 0, 0), (400, 0, 0), and (200, 400, 0) m. The initial positions of the drone users are (-300, 250, 100), (-250, 150, 100), (500, 300, 100), (550, 200, 100), and (600, 100, 100) m, and the initial velocities are (15, 0, 0), (20, 0, 0), (-10, 0, 0), (-15, 0, 0), and (-20, 0, 0) m / s. The state noise is set to... , The measured noise figures are respectively Its trajectory is as follows Figure 3 As shown.
[0037] In step three, each base station transmits the updated state estimate and covariance information to the CPU via the backhaul link for information fusion, and derives explicit analytical expressions for the predictive posterior Cramer-Rao bound (PCRB) of the tracking error and the uplink pilot length: Set the first The state estimates and covariances of each drone user are: and The CPU uses the covariance cross criterion for information fusion. ; ; in The fusion coefficients reflect the confidence level of the corresponding local posterior estimate information and satisfy the following: ; Then, the predictive PCRB considering the information fusion process is derived. for: ; in Let be the Fisher information matrix of the previous time slot, and be the reciprocal of PCRB. Defined based on the relationship between the measurement noise covariance and the uplink pilot length. , ,in for: ; in The noise proportionality coefficients were measured for time delay, Doppler, azimuth, and elevation angle parameters, respectively; then... Eigenvalue decomposition yields: ; in The eigenvector matrix, The matrix is a diagonal matrix composed of eigenvalues; the derivation yields: ; in for The One element, for diagonal Each element.
[0038] Step four: Based on the analytical expression obtained in step three, the CPU constructs a resource allocation problem with the goal of maximizing efficiency and spectral efficiency, and with the constraints of estimation error accuracy requirements and power-association limitations. A suitable combinatorial optimization algorithm is selected to solve the problem and obtain the optimal uplink pilot length, downlink base station user association scheme, and power allocation strategy.
[0039] The resource allocation issue at the CPU level is as follows: ; in , Represents the set of related variables and power allocation variables. To display the tracking error metric defined by the analytical expression between the predicted PCRB obtained in step three and the uplink pilot length: ; To meet tracking error requirements, It is a set of positive integers. and These are the maximum number of users that each base station can serve and the maximum transmission power, respectively. Since only the tracking error constraint in this resource allocation optimization problem is related to the pilot length, and the pilot length has a monotonic relationship with the effective and spectral efficiencies, the optimal uplink pilot length can be obtained first by solving the tracking error metric expression. After fixing the optimal pilot length, the remaining problem is a mixed-integer non-convex optimization problem, which can be solved iteratively with low complexity using compressed sensing and fractional programming. This is achieved by introducing a compressed sensing relaxation correlation scalar: ; in The zero norm represents the number of non-zero values. As a weighting factor, it is updated using the following iterative method: ; in Indicates the number of iterations. As a smoothing factor, it ensures the stability of numerical computation in each iteration. The core steps of solving using fractional programming in each iteration are: ; ; ; in , and These are, respectively, the signal-to-interference-plus-noise ratio (SINR) auxiliary variable, the quadratic transformation auxiliary variable in fractional programming, and the power update variable. For the effective and spectral efficiency coefficients, where The optimal uplink pilot length is obtained by solving the expression for the tracking error metric. and These are the Lagrange multipliers corresponding to the user association constraints and power constraints of the base station; Therefore, the optimal uplink pilot length is first calculated using the tracking error metric expression. Then update sequentially using an iterative method. , , The optimal resource allocation result is obtained when the objective function converges.
[0040] In this embodiment of the invention, the uplink user transmit power is set to... The maximum number of users that each base station can serve is set to The maximum downlink transmit power is In the embodiments, the uplink and downlink power are kept consistent so that the effectiveness and spectral efficiency of different schemes can be compared fairly.
[0041] In step five, the CPU sends the fused state estimate and covariance from step three, as well as the resource allocation results from step four, to each base station. The base station performs state estimation and covariance prediction, and performs predictive beamforming to complete downlink data transmission.
[0042] No. After obtaining the fusion information and resource allocation results, the first base station first predicts the th base station based on the state equation. Status prediction for individual drone users Then, the channel is reconstructed using a line-of-sight-dominated channel model based on the location of the ground base station. And thus obtain the first Reconstructed channel matrix for all users at each base station To reduce inter-user interference, zero-forcing precoding is used to achieve predictive transmit beamforming. : ; in No. The base station for the first The transmit beamforming vector of each UAV; then the base station side based on Perform downlink data transmission until the next beam tracking frame, then repeat steps two through five.
[0043] This invention evaluates a predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion through simulation experiments. Utilizing... Figure 2 In (a), the traditional frame structure does not include information fusion, and resource allocation adopts a fixed pilot length L=16. Uniform power allocation is a traditional method. By comparing the trajectory tracking and position tracking errors, as well as the effective and spectral efficiency results of the traditional method and the low-overhead cooperative beam tracking scheme proposed in this invention, it can be found that the method proposed in this invention achieves better tracking and communication performance than the traditional method while significantly reducing training overhead.
[0044] Figure 3 The system scene settings and trajectory tracking results are shown. It can be seen that the method proposed in this invention achieves tracking accuracy comparable to traditional methods, while requiring only approximately the same level of precision as traditional methods. N c One-third of the training cost. Figure 4 The results further demonstrate the changes in root mean square error (RMSE) and PCRB of position tracking over observation time. PCRB, as the lower limit of RMSE, verifies the correctness of the derivation in step three. Simultaneously, because the method proposed in this invention utilizes… Figure 2 (b) In the beam tracking frame structure, beam training and state estimation are performed only in the first time slot of each frame; prediction is performed only in the remaining time slots. Figure 4 The RMSE and PCRB values corresponding to the method proposed in this invention both exhibited periodic sawtooth patterns due to error accumulation. Finally, it is noted that the method proposed in this invention achieved a lower position tracking error than the traditional method within the observation time.
[0045] Figure 5 The diagram shows the results of the communication system's efficiency and spectral efficiency as a function of the maximum transmit power at the base station. It can be observed that the proposed solution outperforms traditional methods in both efficiency and spectral efficiency due to its lower position tracking error and optimized resource allocation. Furthermore, as the maximum transmit power increases and inter-user interference intensifies, the performance gain of the proposed solution compared to traditional methods becomes even more significant.
[0046] Those skilled in the art can adaptively modify the modules in the embodiments and set them in optimized methods or devices different from those in this embodiment. Specifically, multiple modules in the embodiments can be combined into one module, or a module can be divided into multiple sub-modules and applied to methods or devices with the same technical concept as this embodiment.
Claims
1. A predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion, characterized in that, The method includes the following steps: Step 1: Establish uplink beam training and downlink data transmission models for a cellular-free massive MIMO-assisted UAV communication system, and construct downlink efficiency and spectral efficiency models; Step 2: Using the proposed beam tracking frame structure, the base station uses the uplink pilot signal as a measurement based on the extended Kalman filter to update the UAV user state estimation and covariance information; Step 3: Each base station transmits the updated state estimate and covariance information to the CPU via the backhaul link for information fusion, and derives the explicit analytical expressions for the predictive posterior Cramer-Rao bound (PCRB) of the tracking error and the uplink pilot length. Step 4: Based on the analytical expression obtained in Step 3, the CPU constructs a resource allocation problem with the goal of maximizing efficiency and spectral efficiency, and with the requirements of estimation error accuracy and power-association constraints as conditions. It uses compressed sensing and fractional programming to solve the problem iteratively with low complexity, and obtains the optimal uplink pilot length, downlink base station user association scheme and power allocation strategy. Step 5: The CPU sends the fused state estimate and covariance from Step 3, as well as the resource allocation results from Step 4, to each base station. The base station performs state estimation and covariance prediction, and performs predictive beamforming to complete downlink data transmission.
2. The predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion according to claim 1, characterized in that, Step one specifically includes the following steps: Within the area of interest, there are One ground base station The service is provided to each drone user, and the number of array elements deployed at each ground base station is as follows: A uniform planar array, wherein and These represent the number of antennas in the horizontal and vertical directions, respectively; each drone user has a single antenna. The ground base station is connected to the CPU via a wireless backhaul link for centralized resource scheduling. In the time slot The length sent by the drone user is Orthogonal pilot signal Give AP, number Uplink received signal from each base station for: ； in For the continuous time variable of the uplink received signal, For the first Uplink transmit power of each drone user For the first The base station and the first The communication channels between individual drone users are dominated by line-of-sight. Transmit pulse shaping signal, For the first The base station and the first latency between drone users The base station side uses Gaussian white noise; by projecting the uplink received signal onto different orthogonal pilots, different drone user received signals are obtained. for: ； in This refers to the received noise after projection. Next, consider the downlink data transmission phase, in the time slot. , No. Downlink signal received at a drone user's location for: ； in , and Representing the first The base station for the first The variables associated with each UAV include downlink transmit power and transmit beamforming vector. Through the predictive beamforming construction in step five, For unit power downlink transmission signals, The user-side Gaussian white noise is used to obtain the effective and spectral efficiency of the communication system. for: ； in and Each time slot and symbol time are represented separately. This represents the noise power on the user side.
3. The predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion according to claim 2, characterized in that, Step two specifically includes the following steps: A beam tracking frame structure is adopted, and each frame of the beam tracking frame structure contains In one time slot, the drone user only needs to perform uplink beam training in the first time slot. The base station uses the uplink pilot received signal as a measurement based on the extended Kalman filter to realize the update of drone user state estimation and covariance information. Let the first The state vector of each drone user is: ,in and For drone users in time slots The position and velocity components are obtained by parameter estimation of the projected uplink pilot received signal using the maximum likelihood method. The base station regarding the first Measurement vectors of drone users ; Distributed beam tracking is performed using extended Kalman filtering at each base station; with the first Taking drone user status tracking as an example, the core steps of extended Kalman filtering include two stages: prediction and update. The prediction process includes state estimation and covariance prediction: ； ； in and These are the state estimate and covariance matrix after information fusion, respectively. and These are the predicted state estimate and the covariance matrix, respectively. The motion state transition equation for the UAV is as follows: The state noise covariance; The update process includes Kalman gain calculation and state estimation covariance update: ； ； ； in For Kalman gain, For the measurement equation, To measure the linearization matrix of the equation, To measure the noise covariance, and The first The measurement at the first base station obtained the first Individual drone user state estimation and covariance matrix update It is an identity matrix.
4. The predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion according to claim 3, characterized in that, Step three specifically includes the following steps: Set the first The state estimates and covariances of each drone user are: and The CPU uses the covariance cross criterion for information fusion. ； ； in The fusion coefficients reflect the confidence level of the corresponding local posterior estimate information and satisfy the following: ； Then, the predictive PCRB considering the information fusion process is derived. for: ； in Let be the Fisher information matrix of the previous time slot, and be the reciprocal of PCRB. Defined based on the relationship between the measurement noise covariance and the uplink pilot length. , ,in for: ； in The noise proportionality coefficients were measured for time delay, Doppler, azimuth, and elevation angle parameters, respectively; then... Eigenvalue decomposition yields: ； in The eigenvector matrix, The matrix is a diagonal matrix composed of eigenvalues; the derivation yields: ； in for The One element, for diagonal Each element.
5. The predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion as described in claim 4, characterized in that, Step four specifically includes the following steps: The resource allocation issue at the CPU level is as follows: ； in , Represents the set of related variables and power allocation variables. To display the tracking error metric defined by the analytical expression between the predicted PCRB obtained in step three and the uplink pilot length: ； To meet tracking error requirements, It is a set of positive integers. and These are the maximum number of users that each base station can serve and the maximum transmission power, respectively. First, the optimal uplink pilot length is obtained by solving the tracking error metric expression. After fixing the optimal pilot length, the remaining problem is a mixed-integer non-convex optimization problem, which is solved iteratively with low complexity using compressed sensing and fractional programming. By introducing a compressed sensing relaxation correlation scalar: ； in The zero norm represents the number of non-zero values. As a weighting factor, it is updated using the following iterative method: ； in Indicates the number of iterations. As a smoothing factor, it ensures the stability of numerical computation in each iteration; the core steps of solving using fractional programming in each iteration are: ； ； ； in , and These are, respectively, the signal-to-interference-plus-noise ratio (SINR) auxiliary variable, the quadratic transformation auxiliary variable in fractional programming, and the power update variable. For the effective and spectral efficiency coefficients, where The optimal uplink pilot length is obtained by solving the expression for the tracking error metric. and These are the Lagrange multipliers corresponding to the user association constraints and power constraints of the base station; Therefore, the optimal uplink pilot length is first calculated using the tracking error metric expression. Then update sequentially using an iterative method. , , The optimal resource allocation result is obtained when the objective function converges.
6. The predictive cooperative beamforming and resource allocation method based on non-cellular multi-site information fusion as described in claim 5, characterized in that, Step five specifically includes the following steps: No. After obtaining the fusion information and resource allocation results, the first base station first predicts the th base station based on the state equation. Status prediction for individual drone users Then, the channel is reconstructed using a line-of-sight-dominated channel model based on the location of the ground base station. And thus obtain the first Reconstructed channel matrix for all users at each base station To reduce inter-user interference, zero-forcing precoding is used to achieve predictive transmit beamforming. : ； in No. The base station for the first The transmit beamforming vector of each UAV; then the base station side based on Perform downlink data transmission until the next beam tracking frame, then repeat steps two through five.