A method for mobile perception prediction beamforming

Abstract

The present invention discloses a method for mobile sensing prediction beamforming, which relates to wireless communication. This solution is proposed in view of the problem that the existing beamforming is not efficient enough. Steps: S1. Construct a new sensing-assisted SWIPT system model; S2. Construct a sensing prediction and tracking model of the motion state of a multi-antenna hybrid access point HAP for a mobile energy user MEU; S3. Construct an online beamforming framework based on the extended Kalman filter EKF; S4. Construct an optimization problem model that maximizes the ratio of the power collected by the mobile energy user MEU to the total power consumption at the multi-antenna hybrid access point HAP; S5. Based on the Dinkelbach and SDR transforms, perform convex optimization on the non-convex optimization problem to maximize the wireless power transfer WPT efficiency of the sensing-assisted SWIPT system. The advantage is that the multi-antenna hybrid access point HAP of the constructed sensing-assisted SWIPT system serves multiple single-antenna communication users IU and a mobile energy user MEU, which is closer to the actual application scenario and realizes accurate tracking accuracy and efficient power transmission.

Description

Technical Field

[0001] The present invention relates to wireless communication technologies, and in particular, to a method for mobile sensing prediction beamforming. Background Art

[0002] With the increasing popularity of the Internet of Things, a large number of emerging Internet of Things applications and scenarios, such as unmanned factories and smart homes, have begun to attract public attention. These scenarios usually require the deployment of a large number of low-power sensors to achieve sustainable environmental monitoring, data collection, and intelligent control. As a promising technology, simultaneous wireless information and power transfer (SWIPT) has become one of the most promising solutions, enabling a base station (BS) or an access point (AP) to provide both information transmission and wireless charging simultaneously. In particular, sensors can obtain energy from remote sources via radio frequency (RF), enabling true wireless power supply, thus opening up new ways for the wide deployment and sustainable operation of Internet of Things devices.

[0003] Due to emerging applications of sensors installed on moving objects, such as unmanned transport vehicles and inspection robots, wireless energy transfer faces a series of new challenges in dynamic environments. Especially in highly dynamic scenarios, obtaining accurate sensor position information becomes crucial. However, the frequent acquisition of channel state information (CSI) exacerbates the energy loss caused by the above feedback mechanism. Therefore, it is crucial to achieve timely sensing and positioning of mobile sensors.

[0004] Communication sensing integrated access and control (ISAC) enables a base station to wirelessly sense and characterize the environment and has become a key technology for future 6G networks. Recent research has demonstrated that ISAC can simplify high-dynamic wireless communication, thereby bypassing the cumbersome feedback mechanism and providing a possible solution for the positioning of mobile sensors, and significantly improving the wireless power transfer (WPT) efficiency in SWIPT sensing-assisted SWIPT systems.

[0005] However, existing technologies are limited to beamforming design in static scenarios, neither considering moving energy receivers nor specific channel state information (CSI) acquisition. Once the receiver deviates from a stable stationary state, frequent CSI feedback will result in a reduction in the actually collected energy. Therefore, how to design an efficient beamforming algorithm to ensure real-time and accurate prediction and tracking of mobile energy users (MEUs) while meeting the requirements of information transmission and energy collection is still a challenge faced by SWIPT technology in practical applications. Summary of the Invention

[0006] The object of the present invention is to provide a method for mobile sensing prediction beamforming to solve the problems existing in the above-mentioned prior art.

[0007] The method of mobile sensing prediction beamforming in the present invention includes the following steps:

[0008] S1. Construct a new sensing-assisted SWIPT system model;

[0009] S2. Construct a sensing prediction and tracking model of the motion state of a mobile energy user (MEU) by a multi-antenna hybrid access point (HAP);

[0010] S3. Construct an online beamforming framework based on the extended Kalman filter (EKF);

[0011] S4. Construct an optimization problem model that maximizes the ratio of the power of the mobile energy user (MEU) for energy harvesting to the total power consumption at the multi-antenna hybrid access point (HAP);

[0012] S5. Use Dinkelbach and SDR transformations to perform convex optimization on the non-convex optimization problem, and maximize the wireless power transfer (WPT) efficiency of the sensing-assisted SWIPT system.

[0013] The method of mobile sensing prediction beamforming in the present invention has the advantages that the multi-antenna hybrid access point (HAP) of the constructed sensing-assisted SWIPT system serves multiple single-antenna communication users (IUs) and a mobile energy user (MEU), which is closer to the actual application scenario. Beamforming is performed based on the dynamic prediction of the extended Kalman filter (EKF), the dynamic mobile energy user (MEU) is tracked using the echo signal, and the channel state information (CSI) is inferred by predicting the motion parameters of the mobile energy user (MEU), achieving accurate tracking accuracy and efficient power transmission. The beamforming is formulated as a problem of maximizing the wireless power transfer (WPT) efficiency, while meeting the quality of service requirements of both the single-antenna communication users (IUs) and the mobile energy user (MEU). By applying Dinkelbach and SDR transformations, the original non-convex optimization problem is simplified and effectively solved, greatly reducing the computational complexity and meeting the real-time processing requirements of the communication sensing-assisted SWIPT system. Description of the Drawings

[0014] Figure 1 is a schematic flow chart of the method of mobile sensing prediction beamforming in the present invention.

[0015] Figure 2 is a schematic structural diagram of the sensing-assisted SWIPT system model in the present invention.

[0016] Figure 3 is a schematic flow chart of the beamforming algorithm in the present invention.

[0017] Figure 4It is a schematic diagram of the geometric distribution of a single-antenna communication user IU and a multi-antenna hybrid access point HAP in a single motion state.

[0018] Figure 5 It is a schematic diagram of the geometric distribution of a single-antenna communication user IU and a multi-antenna hybrid access point HAP in multiple motion states.

[0019] Figure 6 It is a schematic diagram of the real-time estimated position of a mobile energy user MEU.

[0020] Figure 7 It is a schematic diagram when the real-time estimated position of the mobile energy user MEU is decomposed into independent X-axis components.

[0021] Figure 8 It is a schematic diagram when the real-time estimated position of the mobile energy user MEU is decomposed into independent Y-axis components.

[0022] Figure 9 It is a real-time energy transfer efficiency curve graph of the method described in the present invention under different tracking upper limits.

[0023] Figure 10 It is a real-time average energy transfer efficiency curve graph of the method described in the present invention under different tracking upper limits. Detailed implementation manners

[0024] A method for mobile perception prediction beamforming described in the present invention, as Figures 1 to 3 shown, includes the following steps:

[0025] S1. Construct a new perception-assisted SWIPT system model.

[0026] Construct a perception-assisted SWIPT system as Figure 2 shown, including a multi-antenna hybrid access point HAP sending communication signals to K single-antenna communication users IU, where the serial number of the single-antenna communication user IU is represented as k ∈ K = {1,..., K}. At the same time, the multi-antenna hybrid access point HAP provides wireless charging for a point-like mobile energy user MEU moving within a certain reachable range. The multi-antenna hybrid access point HAP is equipped with a uniform linear array, including M t transmitting antennas and M r receiving antennas. The mobile energy user MEU is equipped with a radio frequency energy harvesting module and moves along a preset route unknown to the multi-antenna hybrid access point HAP within the total mission time T. The multi-antenna hybrid access point HAP is fixed at a known position, which is represented in a two-dimensional coordinate system as: The single-antenna communication user IU is fixed at another known position, which is represented in a two-dimensional coordinate system as:

[0027] Among them, represents the position of the multi-antenna hybrid access point HAP on the x-axis, represents the position of the multi-antenna hybrid access point HAP on the y-axis; x k represents the position of the k-th single-antenna communication user IU on the x-axis, and y k represents the position of the k-th single-antenna communication user IU on the y-axis; the superscript (·) T is defined as taking the transpose operation, and the value range of k is k ∈ {1, …, K}, where K < M t .

[0028] S2. Construct a perception prediction and tracking model of the motion state of the mobile energy user MEU by the multi-antenna hybrid access point HAP.

[0029] Divide the service time T of the multi-antenna hybrid access point HAP into L equal lengths δ, that is, T = Lδ. In each frame, the multi-antenna hybrid access point HAP first predicts the motion parameters of the mobile energy user MEU, and then directly infers the channel state information CSI according to the motion parameters.

[0030] The channel in the sensing-assisted SWIPT system is a line-of-sight channel, so it follows the free space path loss model. The channel from the multi-antenna hybrid access point HAP to the k-th single-antenna communication user IU in the l-th frame is expressed as: The channel from the multi-antenna hybrid access point HAP to the mobile energy user MEU in the l-th frame is expressed as:

[0031] β0 is the channel gain at the reference distance, d k is the Euclidean distance between the multi-antenna hybrid access point HAP and the k-th single-antenna communication user IU, d E,l is the Euclidean distance between the multi-antenna hybrid access point HAP and the mobile energy user MEU, is the azimuth angle between the multi-antenna hybrid access point HAP and the k-th single-antenna communication user IU, is the azimuth angle between the multi-antenna hybrid access point HAP and the mobile energy user MEU. The transmit array steering vector is modeled as:

[0032]

[0033] represents the azimuth angle.

[0034] In the l-th frame, the downlink signal transmitted by the multi-antenna hybrid access point HAP is represented by and is specifically expressed as: Among them, represents the space of M t ×1 complex number matrices, w k,l is the beamforming vector of the k-th single-antenna communication user IU, v l is the beamforming vector of the mobile energy user MEU, and the information-bearing signal is expressed as denotes obeys a normal distribution with a mean of 0 and a variance of 1. denotes any element of the symbol set K. And the dedicated energy-bearing signal is generated by an arbitrary distribution with E{|s0| 2} = 1.

[0035] The total transmit power P t should not exceed the power budget P of the sensing-assisted SWIPT system max , that is:

[0036]

[0037] where ‖·‖ represents the norm of a vector.

[0038] Therefore, the received signal at the k-th single-antenna communication user IU in the l-th frame is expressed as:

[0039]

[0040] where z k ~CN(0,σ k 2 ) is additive white Gaussian noise and obeys a normal distribution with a mean of 0 and a variance of σ k 2 .

[0041] The received signal-to-interference-plus-noise ratio γ k,l (w k,l ,v l ) at the k-th single-antenna communication user IU is expressed as:

[0042]

[0043] where |·| represents taking the absolute value.

[0044] For the energy harvesting at the mobile energy user MEU, since all received RF signals are beneficial, the instantaneous RF power at the mobile energy user MEU is expressed as:

[0045]

[0046] where μ ∈ [0, 1] is the energy conversion constant of the sensing-assisted SWIPT system, and the superscript (·) H is defined as the conjugate transpose operation.

[0047] The wireless power transfer (WPT) efficiency of the sensing-assisted SWIPT system is defined as the ratio of the power E collected by the mobile energy user (MEU) to the total power consumption P at the multi-antenna hybrid access point (HAP), tot which is expressed as:

[0048]

[0049] where is the power amplifier efficiency, and P C is the circuit power consumption.

[0050] The multi-antenna hybrid access point (HAP) operates in full-duplex mode. While transmitting the downlink SWIPT signal, it performs radar sensing based on the echo signal r l (t) reflected by the mobile energy user (MEU).

[0051]

[0052] υ l represents the Doppler frequency shift of the echo, τ l represents the time delay of the echo, and z r,l represents the receiver noise of the echo.

[0053] The round-trip channel is modeled as where the reflection coefficient is The radar cross section is denoted as κ, the receiving array steering vector the transmitting array steering vector

[0054]

[0055] The vector represents the estimated parameters obtained by signal processing methods such as matched filtering, and the corresponding estimation variance is inversely proportional to the signal-to-noise ratio Γ l of the received echo at the receiver, which is expressed as:

[0056]

[0057] where α MF is the intermediate frequency gain, and

[0058] The time-varying motion parameters of the mobile energy user (MEU) are defined as

[0059] where Denotes the position of the Mobile Energy User (MEU) in a two-dimensional Cartesian coordinate system.

[0060] Denotes the velocity of the Mobile Energy User (MEU) in a two-dimensional Cartesian coordinate system.

[0061] Therefore, the state transition model from the current frame to the next frame is given by X l = A·X l-1 + z pl where A = [I 2×2 , δ·I 2×2 ; 0 2×2 , I 2×2 represents the linear state transition matrix, I 2×2 is the 2×2 identity matrix, 0 2×2 is the 2×2 zero matrix, z pl is the process noise vector, z pl ~ CN(0, Q Z ), z pl follows a complex Gaussian distribution with mean 0 and covariance Q Z , and Q Z is the process noise.

[0062] S3. Construct an online beamforming framework based on the Extended Kalman Filter (EKF).

[0063] To achieve closed-loop target prediction and tracking, a framework based on the Extended Kalman Filter (EKF) is adopted. The Extended Kalman Filter (EKF) is an effective state estimation method that can fuse measurements from multiple sensors to reduce uncertainty and errors. In addition, the radar observation model that characterizes the relationship between the estimated parameters and the motion parameters is:

[0064] O l = f(X l ) + z sl ;

[0065] where f(·) is the non-linear observation operator, z sl is the radar estimation noise vector, z sl ~ CN(0, Q S ), and the estimated noise covariance matrix

[0066] The fusion steps for the estimated and predicted data are as follows:

[0067] At the beginning of each frame, the state prediction is The predicted covariance matrix The superscript denotes the variable based on prediction, and the superscript ~ denotes the variable corrected after prediction and sensing fusion. Subsequently, the Kalman filter gain is calculated as where denotes the Jacobian matrix of the observation model. diag(x) is defined as a diagonal matrix where each diagonal element is the corresponding element in x. Therefore, the updated motion state is expressed as:

[0068] The predicted covariance matrix is expressed as:

[0069] By iteratively executing and updating the motion state and the predicted covariance matrix M l , the multi-antenna hybrid access point HAP can achieve the positioning and tracking of the mobile energy user MEU.

[0070] S4. Construct an optimization problem model that maximizes the ratio of the power collected by the mobile energy user MEU to the total power consumption at the multi-antenna hybrid access point HAP.

[0071] The objective of this embodiment is to maximize the real-time wireless power transfer WPT efficiency per frame through optimizing the transmit beamforming, while meeting the quality of service requirements of the sensing-assisted SWIPT system, including the signal-to-interference-plus-noise ratio requirements of the single-antenna communication user IU and the energy harvesting and sensing quality of service of the mobile energy user MEU.

[0072] The trace of the predicted covariance matrix represents the mean square error after state tracking. Therefore, the instantaneous tracking performance is defined as S = tr(M l ), which is an implicit function of the beamforming vector v, and tr(·) is defined as the trace of a matrix. The corresponding beamforming problem is formulated as follows:

[0073]

[0074] s.t. P(w k , v) ≤ P max ,

[0075] SINR k (w k ) ≥ γ min ,

[0076]

[0077] S(v) ≤ C max .

[0078] where, s.t. is defined as subject to; P(w k , v) ≤ P max represents the transmit power budget; SINR k (w k ) ≥ γ minGuarantee the minimum signal-to-interference-plus-noise ratio (SINR) for single-antenna communication users IU; for the harvested energy constraint, Denote the predicted energy under the constraint of beamforming while maintaining SINR k (w k ) ≥ γ min as e min Denote the minimum power requirement at the l-th frame; the predefined parameter C max Denote the upper bound of the tracking performance.

[0079] Combine the estimation variance which is inversely proportional to the received signal-to-noise ratio, and further express the updated estimation variance as where the constant a Ω is determined by the sensing-aided SWIPT system configuration and the specific estimation method, and represent the estimated noise covariance matrix as:

[0080]

[0081] S5. Adopt the Dinkelbach and SDR transformation to perform convex optimization on the non-convex optimization problem, and maximize the wireless power transfer (WPT) efficiency of the sensing-aided SWIPT system.

[0082] To solve the non-convex objective function, adopt the Dinkelbach-based non-linear fractional programming, where the objective function in its corresponding subtraction form is given by .

[0083] In the case of using the unique zero solution, transform problem P0 into the following equivalent form:

[0084]

[0085] Introduce the auxiliary variables W = ww H and V = vv H , with the constraint conditions W k ≥ 0, V ≥ 0, rank(W k ) = 1 and rank(V) = 1. rank(·) is defined as the rank of the matrix. Thus, the above optimization problem is rewritten as:

[0086]

[0087] where g represents the predicted channel from the base station to the mobile energy user (MEU). However, due to the rank-one constraint, the problem is still non-convex and can be solved by the classical semi-definite relaxation method. Therefore, simply remove the rank constraint, The problem P1 can be transformed by semidefinite relaxation and effectively solved by a standard solver such as CVX. CVX is developed based on the Matlab environment and is a convex optimization toolbox that can be used to solve problems such as linear programming, quadratic programming, semidefinite programming, and geometric programming. Among them, the feasible rank-one solution of the original problem can be obtained through eigenvalue decomposition or Gaussian randomization procedures.

[0088] The effectiveness of the method proposed in the present invention is demonstrated by the following numerical results.

[0089] The multi-antenna hybrid access point HAP is set at the origin and equipped with 32 transmit antennas and 2 receive antennas to serve 3 single-antenna communication users IU and 1 mobile energy user MEU, that is, M t = 32, M r = 2, K = 3. The sensing-assisted SWIPT system operates in the 5.5 GHz band, and the reference channel power gain is calculated through the free space propagation model. The minimum signal-to-interference-plus-noise ratio required by the single-antenna communication user IU is γ min = 5 dB, and the energy harvesting requirement of the mobile energy user MEU is e min = 1.1 μW. The remaining simulation parameters are set as follows:

[0090] δ = 0.5 s, P max = 40 dBm, σ R = -70 dBm, σ k = -80 dBm, P C = 0.8 W,

[0091] a Ω = [0.085, 0.9×10 6 , 45], α MF = 10 4 , κ = 0.9, η = 0.75,

[0092] Figure 4 and Figure 5 shows the geometric distribution of the single-antenna communication user IU and the multi-antenna hybrid access point HAP, as well as the real-time tracking results based on the proposed scheme under different target movements.

[0093] In the process of the mobile energy user MEU in a single motion state, the process noise z pl is small, while in the multiple motion states, the process noise experienced by the mobile energy user MEU is quite large. It can be observed that the real-time tracking trajectories generally match the actual curves in different motion cases, demonstrating that the present invention has good tracking performance.

[0094] In a multi - motion state, when the mobile energy user (MEU) experiences a sharp turn that causes a sudden increase in process noise, trajectory deviations are observed. However, after several frames of real - time tracking adjustment, these deviations gradually decrease, enabling the trajectory to quickly realign with the actual path.

[0095] To further demonstrate the performance of the present invention, the settings in the multi - motion state are followed and the tracking performance of the instantaneous position mean - square error over the entire service time is evaluated. As Figure 6 shown, over the entire operation time, the present invention maintains a lower mean - square error compared to the prior art. In addition, this improvement gap is very obvious around the 22nd and 44th frames because the mobile energy user (MEU) experiences turns that the state - transition model cannot accurately model. However, the solution of the present invention significantly reduces the positioning error and exhibits superior tracking performance. This positioning improvement is more clearly illustrated in Figure 7 and Figure 8 where the position of the mobile energy user (MEU) is decomposed into independent X - and Y - axes. The results show that the present invention is closer to the actual trajectory compared to the prior art without sensing assistance.

[0096] The real - time wireless power transfer (WPT) efficiency of the present invention under different tracking upper limits is as shown in Figure 9 and Figure 10 The UB scheme represents the ideal optimum based on an accurate prior of the mobile energy user (MEU) state and requires no resource consumption. These three distinct fluctuating parts reflect the motion state of the mobile energy user (MEU) because the radio - frequency signal strength decreases with the increase of distance d E,l and the three peaks exactly correspond to the moments when d E,l reaches the minimum value. The present invention is close to the ideal upper bound in overall performance, but there are obvious gaps and fluctuations due to real - time positioning errors caused by sudden changes in the motion of the mobile energy user (MEU).

[0097] As the constraint value increases, the performance degradation becomes more obvious. This is due to insufficient sensing accuracy, resulting in obvious channel prediction errors in subsequent frames, thus reducing the performance of the sensing - assisted SWIPT system.

[0098] The results also reveal that a more stringent sensing - accuracy constraint does not necessarily lead to higher wireless power transfer (WPT) efficiency because it consumes too many resources of the sensing - assisted SWIPT system.

[0099] As shown above, the method of mobile sensing prediction beamforming in the present invention can achieve efficient power transmission. Beamforming is formulated as a problem of maximizing the efficiency of wireless power transfer (WPT), while satisfying the quality of service requirements of single-antenna communication users (IUs) and mobile energy users (MEUs). By applying the Dinkelbach and semidefinite relaxation methods, the original non-convex optimization problem is simplified and effectively solved. Simulation results are provided to verify the superiority of the proposed scheme in improving the performance of the sensing-assisted SWIPT system.

[0100] For those skilled in the art, according to the technical solutions and concepts described above, various corresponding changes and deformations can be made, and all these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A method for mobile perception prediction beamforming, characterized in that It includes the following steps: S1. Construct a new perception-assisted SWIPT system model; S2. Construct a perception prediction and tracking model for the motion state of the mobile energy user (MEU) by the multi-antenna hybrid access point (HAP); S3. Construct an online beamforming framework based on the extended Kalman filter (EKF); S4. Construct an optimization problem model that maximizes the ratio of the power of the mobile energy user (MEU) for energy harvesting to the total power consumption at the multi-antenna hybrid access point (HAP); S5. Use Dinkelbach and SDR transformation to perform convex optimization on the non-convex optimization problem to maximize the wireless power transfer (WPT) efficiency of the perception-assisted SWIPT system; In the step S1, the constructed sensing-assisted SWIPT system includes a multi-antenna hybrid access point HAP that transmits communication signals to single-antenna communication users IU, where the serial number of the single-antenna communication user IU is represented as ; the multi-antenna hybrid access point HAP provides wireless charging for a point-like mobile energy user MEU that moves within a certain reachable range; the multi-antenna hybrid access point HAP is equipped with a uniform linear array, including transmitting antennas and receiving antennas; the mobile energy user MEU is equipped with a radio frequency energy harvesting module and moves along a preset route unknown to the multi-antenna hybrid access point HAP within the total mission time . The multi-antenna hybrid access point HAP is fixed at a known position and is represented in a two-dimensional coordinate system as: ; The single-antenna communication user IU is fixed at another known position, represented in a two-dimensional coordinate system as: ： Among them, represents the position of the multi-antenna hybrid access point HAP on the x-axis, represents the position of the multi-antenna hybrid access point HAP on the y-axis; represents the th position of the single-antenna communication user IU on the x-axis, represents the th position of the single-antenna communication user IU on the y-axis; The superscript is defined as taking the transpose operation, The value range of is ; In the step S2, the service time of the multi-antenna hybrid access point HAP is divided into equal lengths , that is ; within each frame, the multi-antenna hybrid access point HAP first predicts the motion parameters of the mobile energy user MEU, and then directly infers the channel state information CSI according to the motion parameters; The channel in the perception-assisted SWIPT system is a line-of-sight channel and follows the free space path loss model.

2. The method for mobile perception prediction beamforming according to claim 1, characterized in that The channel from the multi-antenna hybrid access point HAP to the single-antenna communication user IU in the ； The channel from the multi-antenna hybrid access point HAP to the mobile energy user MEU in the ； is the channel gain at the reference distance, is the Euclidean distance between the multi-antenna hybrid access point HAP and the th single-antenna communication user IU, is the Euclidean distance between the multi-antenna hybrid access point HAP and the mobile energy user MEU, is the azimuth angle between the multi-antenna hybrid access point HAP and the th single-antenna communication user IU, is the azimuth angle between the multi-antenna hybrid access point HAP and the mobile energy user MEU; The transmit array steering vector is modeled as: , Indicates the azimuth angle; At the th frame, the downlink signal transmitted by the multi-antenna hybrid access point HAP is expressed as: ； Among them, is the beamforming vector of the th single-antenna communication user IU, is the beamforming vector of the mobile energy user MEU, and the information-bearing signal is expressed as ; denote obeys a normal distribution with a mean of 0 and a variance of 1, denote the symbol set any element of; while the dedicated energy-bearing signal is generated by any distribution of Total transmit power does not exceed the power budget of the sensing-aided SWIPT system : ； Among them, represents the norm of a vector; The received signal at the -th single-antenna communication user IU in the -th frame is expressed as: ； wherein is additive white Gaussian noise, and follows a normal distribution with a mean of 0 and a variance of ; the received signal-to-interference-plus-noise ratio at the $i$-th single-antenna communication user IU is expressed as: as follows: ； Among them, represents taking the absolute value; The instantaneous RF power at the mobile energy user (MEU) is expressed as: ； Among them is the energy conversion constant of the perception-assisted SWIPT system, and the superscript is defined as the conjugate transpose operation; The wireless power transfer (WPT) efficiency of the perception-assisted SWIPT system is expressed as: ； Wherein, , is the power amplifier efficiency, is the circuit power consumption; The multi-antenna hybrid access point (HAP) operates in full-duplex mode and performs radar sensing based on the echo signal reflected by the mobile energy user (MEU) while transmitting the downlink SWIPT signal. Perform radar sensing; ； represents the Doppler shift of the echo, represents the time delay of the echo, represents the receiver noise of the echo; The round-trip channel is modeled , where the reflection coefficient is , the radar cross section is expressed as , the receiving array steering vector , the transmitting array steering vector , the vector represents the estimated parameters obtained by signal processing methods such as matched filtering, and the corresponding estimated variance , , is inversely proportional to the signal-to-noise ratio of the received echo and is expressed by the following formula: ； wherein is the intermediate frequency gain, is the noise of the multi-antenna hybrid access point HAP receiver; Define the time-varying motion parameters of the mobile energy user MEU as ; wherein, represents the position of the Mobile Energy User (MEU) in a two-dimensional Cartesian coordinate system, represents the velocity of the Mobile Energy User (MEU) in a two-dimensional Cartesian coordinate system; The state transition model from the current frame to the next frame is given by where denotes the linear state transition matrix, is the 2×2 identity matrix, is the 2×2 zero matrix, is the process noise vector, , obeys a complex Gaussian distribution with mean 0 and covariance , is the process noise.

3. The method for mobile sensing prediction beamforming according to claim 2, wherein In step S3, the radar observation model that uses the framework based on the extended Kalman filter (EKF) to characterize the relationship between the estimated parameters and the motion parameters is: ； In the formula, is a non - linear observation operator, is the radar estimation noise vector, , where the estimation noise covariance matrix ; The fusion steps of the estimated and predicted data are as follows: At the start of each frame, the state is predicted as , the predicted covariance matrix , the superscript denotes the variable based on prediction, and the superscript denotes the variable corrected after prediction and sensor fusion; subsequently, the Kalman filter gain is calculated as , where represents the Jacobian matrix of the observation model; is defined as a diagonal matrix whose each diagonal element is the corresponding element in ; The updated motion state is represented as: ; The predicted covariance matrix is expressed as: ; By iteratively executing and updating the motion state and the predicted covariance matrix , the multi-antenna hybrid access point HAP realizes the positioning and tracking of the mobile energy user MEU.

4. The method for mobile perception prediction beamforming according to claim 3, wherein In the step S4, the trace of the predicted covariance matrix represents the mean square error after state tracking, and the instantaneous tracking performance is defined as , which is an implicit function of the beamforming vector , defined as the trace of the matrix; The corresponding beamforming problem formula is as follows: ； ， ， ， ； Among them, is defined as being constrained by; represents the transmit power budget; ensures the minimum signal-to-interference-plus-noise ratio of the single-antenna communication user IU; for the harvested energy constraint, represents the predicted energy while beamforming under the constraint ; represents the minimum power requirement at the frame; the predefined parameter represents the upper limit of the tracking performance; Combined estimated variance , , A relationship inversely proportional to the received signal-to-noise ratio, the updated expression of the estimated variance is , where , the constant is determined by the sensing-assisted SWIPT system configuration and the specific estimation method, and the estimated noise covariance matrix is expressed as: 。 5. The method for mobile perception prediction beamforming according to claim 4, characterized in that In the step S5, a non-convex objective function is solved by using Dinkelbach-based non-linear fractional programming, where the objective function has a corresponding subtraction form as follows: ； When using the unique zero solution, problem P0 is transformed into the following equivalent form: ； Introduce auxiliary variables and , the constraint conditions are 、 and ; is defined as the rank of the matrix; The optimization problem is rewritten as: ； ， ， ， ， ， ， ， ； where g represents the predicted channel from the base station to the mobile energy user (MEU); Remove the rank constraint, , ; Problem P1 undergoes semi-definite relaxation transformation and is solved by a standard solver.

Images