Reconfigurable intelligent surface energy transfer system for non-line-of-sight environments

Abstract

The application provides a reconfigurable intelligent surface energy transmission system for a non-line-of-sight environment, and belongs to the technical field of wireless power supply systems.The system comprises a feedback module, which is arranged at a mobile receiving end and is used for collecting voltage data on a local energy receiving circuit; a global scanning module, which is used for controlling the RIS to perform beam scanning and determining a maximum receiving power direction as an initial focusing point according to the voltage data; a three-dimensional prediction module, which establishes a three-dimensional voxel grid with the initial focusing point as the center, and estimates the predicted position coordinates of the mobile receiving end by a preset timing prediction model based on historical and real-time voltage data sequences; a control module, which calculates the phase configuration of the RIS unit according to the predicted position coordinates, and synchronously outputs the phase control signal through a shift register to drive the RIS array; and an energy transmission module, which is used for forming a focused beam to the mobile receiving end for wireless power supply according to the phase control signal. The application realizes low-delay and high-precision energy focusing on a mobile target.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of wireless power supply systems, and particularly relates to a reconfigurable intelligent surface energy transmission system for a non-line-of-sight environment. BACKGROUND

[0002] Wireless energy transmission technology makes it possible for Internet of Things devices and mobile robots to operate continuously. However, in real industrial or indoor environments, there are often obstacles between the transmitting end and the receiving end, which blocks the direct line-of-sight (LOS) link, forming a non-line-of-sight (NLOS) environment, so that the traditional energy transmission method based on the line-of-sight path has a sharp decline in efficiency.

[0003] As an emerging technology, the reconfigurable intelligent surface (RIS) can theoretically solve the non-line-of-sight problem by intelligently regulating the wave front of electromagnetic waves to construct a virtual line-of-sight link. However, the existing technology has the following defects: 1. Most existing reconfigurable intelligent surface energy transmission schemes assume that the receiving end is static or can only perform simple beam steering, and cannot effectively track high-speed moving targets in a non-line-of-sight environment. 2. In a non-line-of-sight environment, due to signal blocking and reflection, it is difficult for the system to directly obtain the accurate position of the moving target. Most existing methods rely on instantaneous feedback and lack the ability to predict the target trajectory, resulting in beam pointing always lagging behind the actual position of the target. 3. The performance of the reconfigurable intelligent surface is positively correlated with the number of units, but the control of a large-scale reconfigurable intelligent surface array requires extremely high computing speed and parallel processing capability. The traditional CPU / GPU-based control system has a high processing delay and is difficult to achieve real-time response in nanoseconds. Although FPGA has parallel processing potential, existing designs have deficiencies in driving channel expansion and synchronous control, making it difficult to efficiently control hundreds or even thousands of reconfigurable intelligent surface units. SUMMARY

[0004] To solve the above technical problems, the application provides a reconfigurable intelligent surface energy transmission system for a non-line-of-sight environment. Through the cooperation of real-time feedback by Bluetooth, a three-dimensional motion prediction algorithm, and large-scale array synchronous control based on FPGA, low-delay and high-precision energy focusing and tracking of moving targets are achieved.

[0005] To achieve the above purpose, the application adopts the following technical solutions:

[0006] The reconfigurable intelligent surface energy transmission system for a non-line-of-sight environment comprises

[0007] A feedback module is deployed at the mobile receiving end, which collects voltage data of the local load through a voltage sampling circuit and transmits them to a global scanning module and a three-dimensional prediction module, respectively;

[0008] a global scanning module configured to control the RIS to perform beam scanning and determine a maximum received power direction as an initial focus point according to the voltage data;

[0009] a three-dimensional prediction module in communication with the feedback module and the global scanning module, configured to establish a three-dimensional voxel grid centered on the initial focus point, and estimate a predicted position coordinate of the mobile receiving end based on historical and real-time voltage data sequences continuously returned by the feedback module through a preset time series prediction model;

[0010] a control module in communication with the three-dimensional prediction module, configured to calculate RIS unit phase configurations according to the predicted position coordinate and synchronously output phase control signals through a shift register to drive the RIS array;

[0011] an energy transmission module configured to form a focused beam according to the phase control signals to wirelessly power the mobile receiving end.

[0012] Further, the feedback module comprises:

[0013] an ARM microcontroller configured to collect voltage data of a local load through a voltage sampling circuit;

[0014] a Bluetooth communication unit connected with the ARM microcontroller and configured to return the collected voltage data to the global scanning module and the three-dimensional prediction module, respectively.

[0015] Further, the detailed process of the operation of the global scanning module comprises:

[0016] controlling the RIS array to sequentially load a plurality of groups of phase configurations in a pre-stored beam forming codebook, each group of phase configurations corresponding to a different beam direction;

[0017] traversing all voltage data received during the process of traversing all phase configurations, and determining the beam direction corresponding to the maximum voltage value as the initial focus point.

[0018] Further, the detailed process of the operation of the three-dimensional prediction module comprises:

[0019] establishing a three-dimensional voxel grid centered on the initial focus point;

[0020] at each time step, mapping historical and real-time voltage data sequences from the feedback module to the three-dimensional voxel grid to form an observation tensor, and extracting spatial features through an encoder to obtain a feature tensor;

[0021] inputting the feature tensor into a preset time series prediction model to predict a position probability distribution of the mobile receiving end at a predicted time step;

[0022] Calculate the weighted expectation of the position probability distribution at any time step to obtain the estimated predicted position coordinates.

[0023] Furthermore, the temporal prediction model in the three-dimensional prediction module is a convolutional long short-term memory network, specifically including:

[0024] Input Gate: ;

[0025] Forgotten Gate: ;

[0026] Output gate: ;

[0027] Candidate memories: ;

[0028] Cell status update: ;

[0029] Hidden status update: ;

[0030] in, Index for the current time step; express Time input gate, express The Gate of Forgetting Time; express The output gate at any given time; It is the sigmoid activation function;

[0031] express The convolution kernel weight matrix to the input gate; express The convolution kernel weight matrix leading to the forget gate; express The convolution kernel weight matrix leading to the output gate; express The convolution kernel weight matrix to the candidate state; Indicates the previous hidden state The convolution kernel weight matrix to the input gate; Indicates the previous hidden state The convolution kernel weight matrix leading to the forget gate; Indicates the previous hidden state The convolution kernel weight matrix leading to the output gate; This represents the bias vector of the input gate; The bias vector representing the forget gate; This represents the bias vector of the output gate; The bias vector representing the candidate state; express The time is determined by historical location, received voltage, and RIS status information in a three-dimensional grid. The input feature tensor formed above; express Cellular state at any given moment; Represents the hyperbolic tangent activation function; This indicates element-wise multiplication.

[0032] Furthermore, the training process of the 3D prediction module minimizes the multi-objective loss function, specifically as follows:

[0033] ;

[0034] in, Represents cross-entropy loss; This indicates the position mean square error loss; This indicates a loss of speed consistency. Indicates power regression loss; Indicates the L2 regularization term;

[0035] The weight coefficients representing the cross-entropy loss; The weighting coefficients representing the position mean square error loss; Weighting coefficients representing the speed consistency loss; The weighting coefficients representing the power regression loss; This represents the weight coefficient of the L2 regularization term.

[0036] Furthermore, the process executed by the control module includes:

[0037] Based on the predicted location coordinates Calculate the required phase configuration vector for each cell in the RIS cell array. ;in, This indicates the total number of cells contained in the RIS cell array; Represents the first RIS cell array The phase offset that each unit needs to be configured with;

[0038] Phase configuration vector Each consecutive phase value in Quantized into corresponding digital control words and all Combined into data frames suitable for serial communication;

[0039] The packaged serial data is shifted to a position driven by a clock signal. The shift register chip is cascaded to form a register chain; when all data is in place, a global latch signal is generated to make all shift registers update the data in the internal register to the parallel output end synchronously;

[0040] The level signal output by the shift register is loaded to each corresponding unit of the RIS unit array through the driving circuit, so that all units switch to the new electromagnetic state according to the digital control word , thereby forming a focused beam pointing to .

[0041] Further, the calculation process of the digital control word is as follows:

[0042] ;

[0043] wherein, represents the bit width of the corresponding digital control word for controlling a single RIS unit; represents a function of rounding to the nearest integer.

[0044] Further, the calculation process of the phase offset is as follows:

[0045] ;

[0046] wherein, represents the wavelength of the electromagnetic wave carrier of the system; represents the total propagation path difference of the i-th RIS unit relative to the reference unit to the target point;

[0047] ;

[0048] represents the central position coordinate of the i-th RIS unit; represents the position coordinate of the radio frequency source; represents the position coordinate of the pre-selected reference RIS unit; represents the Euclidean norm operator.

[0049] Further, the energy transmission module includes a radio frequency source and a RIS array:

[0050] The radio frequency source is used to generate a radio frequency signal as an energy carrier;

[0051] The RIS array is used to receive the radio frequency signal, reconstruct the electromagnetic wave front according to the phase control signal issued by the control module, form a focused beam pointing to the predicted position coordinate, and wirelessly power the mobile receiving end.

[0052] ​​The effects provided in the summary are only the effects of the embodiments, not all the full effects of the application, and one of the technical solutions has the following advantages or beneficial effects:

[0053] The application provides a reconfigurable intelligent surface energy transmission system for a non-line-of-sight environment, and belongs to the technical field of wireless power supply systems.

[0054] The application innovatively adopts a ConvLSTM space-time prediction model based on an encoder-decoder architecture, can perform high-precision prediction on the position of a target within hundreds of milliseconds in the future based on historical trajectories and signal strength, and has the advantages of high prediction accuracy and low time delay.

[0055] The application adopts a hardware architecture of "FPGA + cascaded shift register". The FPGA provides nanosecond-level calculation and response speed, and the cascaded circuit based on the SN74HC595 realizes the synchronous driving of hundreds or even thousands of RIS units by using very few FPGA pins through a "serial shift-synchronous latching" mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A reconfigurable intelligent surface energy transmission system for a non-line-of-sight environment is provided for the application example 1;

[0057] Figure 2 A workflow diagram of the global scanning module is provided for the application example 1;

[0058] Figure 3 A workflow diagram of the three-dimensional prediction module is provided for the application example 1;

[0059] Figure 4Grid diagram of the three-dimensional prediction module proposed for the embodiment 1 of the present application;

[0060] Figure 5 Workflow diagram of the control module proposed for the embodiment 1 of the present application;

[0061] Figure 6 Control circuit framework diagram based on 595 proposed for the embodiment 1 of the present application;

[0062] Figure 7 FPGA control timing diagram proposed for the embodiment 1 of the present application. DETAILED DESCRIPTION

[0063] To clearly illustrate the technical features of the present application, the present application is described in detail below through specific embodiments, and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing the different structures of the present application. In order to simplify the disclosure of the present application, the components and settings of specific examples are described below. In addition, the present application can repeatedly refer to numbers and / or letters in different examples. Such repetition is for the purpose of simplification and clarity, and does not in itself indicate the relationship between the various embodiments and / or settings being discussed. It should be noted that the components illustrated in the drawings are not necessarily drawn to scale. The present application omits the description of well-known components and processing techniques and processes to avoid unnecessarily limiting the present application.

[0064] Embodiment 1

[0065] The embodiment 1 of the present application proposes a reconfigurable intelligent surface energy transmission system for non-line-of-sight environment, which is used to solve the technical problem that the transmission end and the receiving end are often in a non-line-of-sight state due to the obstruction of obstacles in the prior art, resulting in a sharp decline in the efficiency of traditional line-of-sight energy transmission methods.

[0066] Figure 1 The schematic diagram of the reconfigurable intelligent surface energy transmission system for non-line-of-sight environment proposed for the embodiment 1 of the present application; the system includes a control center, a feedback module deployed at a mobile receiving end, and a transmitting end including a radio frequency source and a reconfigurable intelligent surface array; the control center includes a global scanning module, a three-dimensional prediction module, and a control module.

[0067] The feedback module is deployed at the mobile receiving end, and is used to collect voltage data on the local energy receiving circuit of the mobile receiving end, and is transmitted back to the global scanning module and the three-dimensional prediction module, respectively;

[0068] The feedback module includes an ARM microcontroller, which collects voltage data of the local load through a voltage sampling circuit; a Bluetooth communication unit connected with the ARM microcontroller, which is used to wirelessly transmit the collected voltage data to the global scanning module and the three-dimensional prediction module, respectively.

[0069] In this invention, the mobile receiver typically employs a mobile vehicle equipped with an ARM microcontroller (such as the STM32F407 series) and an energy receiving circuit (rectifier antenna + load). The ARM microcontroller is responsible for acquiring the voltage values ​​across the load terminals, which are positively correlated with the received power. The acquired data is wirelessly transmitted in real-time to the control center via a Bluetooth module (such as the HC-05 or BLE series). The control center periodically receives the voltage data from the vehicle, using it as a basis for system focusing optimization and prediction.

[0070] The scope of protection of this invention is not limited to the device models listed in Example 1, and those skilled in the art can make reasonable selections based on the actual situation.

[0071] The global scanning module is used to quickly determine the approximate direction of a target during system initialization or when the target is lost.

[0072] The global scanning module, which communicates with the feedback module and the three-dimensional prediction module, is used to control the RIS to perform beam scanning and determine the direction of maximum received power based on voltage data as the initial focusing point.

[0073] Figure 2 This is a schematic diagram of the operation of the global scanning module proposed in Embodiment 1 of the present invention; the detailed process includes: controlling the RIS array to load multiple sets of phase configurations in the pre-stored beamforming codebook in sequence, each set of phase configurations corresponding to a different beam direction; traversing all voltage data received during the process of all phase configurations, determining the RIS phase configuration corresponding to the maximum received voltage as the initial focusing coordinate point, providing a starting point for subsequent three-dimensional fine prediction.

[0074] The 3D prediction module is the core module within the control center, used to predict the future location of moving targets in non-line-of-sight environments.

[0075] The 3D prediction module communicates with the feedback module, global scanning module, and control module. It establishes a 3D voxel grid centered on the initial focus point and estimates the predicted position coordinates of the mobile receiver based on the historical and real-time voltage data sequences continuously transmitted back by the feedback module, using a preset time-series prediction model.

[0076] Figure 3 The flowchart below shows the workflow of the three-dimensional prediction module proposed in Embodiment 1 of this invention: A three-dimensional voxel grid is established with the initial focal point as the center; at each time step, the historical and real-time voltage data sequences from the feedback module are mapped to the three-dimensional voxel grid to form an observation tensor, and spatial features are extracted by the encoder to obtain a feature tensor; the feature tensor is input into a preset time-series prediction model to predict the position probability distribution of the mobile receiver at the prediction time step; the weighted expectation of the position probability distribution at any time step is calculated to obtain the estimated predicted position coordinates.

[0077] Figure 4 This is a schematic diagram of the mesh of the three-dimensional prediction module proposed in Embodiment 1 of the present invention; in this application, a temporal neural network with an encoder-decoder architecture is used, where the encoder E maps the observation tensor composed of historical voltage, RIS phase, and estimated position into a feature tensor. The temporal core employs ConvLSTM (Convolutional Long Short-Term Memory network), and its gating update mechanism is as follows:

[0078] Input Gate: ;

[0079] Forgotten Gate: ;

[0080] Output gate: ;

[0081] Candidate memories: ;

[0082] Cell status update: ;

[0083] Hidden status update: ;

[0084] in, Index for the current time step; express Time input gate, express The Gate of Forgetting Time; express The output gate at any given time; It is the sigmoid activation function;

[0085] express The convolution kernel weight matrix to the input gate; express The convolution kernel weight matrix leading to the forget gate; express The convolution kernel weight matrix leading to the output gate; express The convolution kernel weight matrix to the candidate state; Indicates the previous hidden state The convolution kernel weight matrix to the input gate; Indicates the previous hidden state The convolution kernel weight matrix leading to the forget gate; Indicates the previous hidden state The convolution kernel weight matrix leading to the output gate; This represents the bias vector of the input gate; The bias vector representing the forget gate; This represents the bias vector of the output gate; The bias vector representing the candidate state; express The time is determined by historical location, received voltage, and RIS status information in a three-dimensional grid. The input feature tensor formed above; express Cellular state at any given moment; Represents the hyperbolic tangent activation function; This represents element-wise multiplication;

[0086] Decoder D will output the hidden state of ConvLSTM. The predicted position of the target is obtained by converting the voxel grid prediction score for the next H steps, normalizing it using the Softmax function, and then taking the expected value of this probability distribution.

[0087] The training process of the 3D prediction module minimizes the multi-objective loss function, specifically as follows:

[0088] ;

[0089] in, Represents cross-entropy loss; This indicates the position mean square error loss; This indicates a loss of speed consistency. Indicates power regression loss; Indicates the L2 regularization term;

[0090] The weight coefficients representing the cross-entropy loss; The weighting coefficients representing the position mean square error loss; Weighting coefficients representing the speed consistency loss; The weighting coefficients representing the power regression loss; This represents the weight coefficient of the L2 regularization term.

[0091] The control module, which communicates with the 3D prediction module, is used to calculate the phase configuration of the RIS unit based on the predicted position coordinates and synchronously output the phase control signal through a shift register to drive the RIS array. Figure 5 This is a flowchart of the control module proposed in Embodiment 1 of the present invention; the detailed process includes:

[0092] Based on the predicted location coordinates Calculate the required phase configuration vector for each cell in the RIS cell array. ;in, This indicates the total number of cells contained in the RIS cell array; Represents the first RIS cell array The phase offset that each unit needs to be configured with;

[0093] Phase configuration vector Each consecutive phase value in Quantized into corresponding digital control words and all Combined into data frames suitable for serial communication;

[0094] The packaged serial data is shifted to a position driven by a clock signal. In a register chain composed of cascaded shift register chips, once all data is in place, a global latch signal is generated, causing all shift registers to synchronously update the data in their internal registers to the parallel output.

[0095] The level signal output in parallel from the shift register is driven by the driver circuit and then applied to each corresponding cell in the RIS cell array, so that all cells respond according to their digital control words. Synchronously switch to a new electromagnetic state, thereby forming a directional... The focused beam.

[0096] Digital control word The calculation process is as follows:

[0097] ;

[0098] in, This indicates the bit width of the digital control word that controls a single RIS unit; This function represents rounding down to the nearest integer.

[0099] The calculation process for the phase offset is as follows:

[0100] ;

[0101] in, The electromagnetic carrier wavelength that indicates the system's operation; Indicates the first The total propagation path difference from the reference cell to the target point for each RIS cell;

[0102] ;

[0103] Indicates the first The center coordinates of each RIS unit; Indicates the location coordinates of the radio frequency source; Indicates the position coordinates of a pre-selected reference RIS cell; The operator represents the Euclidean norm.

[0104] In this invention, the control module is the key hardware for achieving low-latency, large-scale control. A Xilinx Artix-7 series FPGA is used as the control core, leveraging its parallel processing capabilities to achieve nanosecond-level phase calculations and command issuance. The scope of protection of this invention is not limited to the models listed in Example 1; those skilled in the art can make reasonable selections based on actual circumstances.

[0105] Figure 6 This invention provides a control circuit framework diagram based on the 595 microcontroller for Embodiment 1. To drive up to 800 RIS units, an SN74HC595PW shift register is used for I / O expansion. Specifically, the FPGA serially inputs control data to the first 595 microcontroller via the SPI communication protocol, and then passes it to the next microcontroller through cascading outputs, forming a long chain of shift registers. Each 595 microcontroller controls 8 RIS units, and 100 microcontrollers can control 800 RIS units. A latch signal is used to simultaneously update the outputs of all 595 microcontrollers, achieving synchronous phase switching of the 800 RIS units. A perforated board circuit is located between the 595 outputs and the RIS units, used for level conversion, signal isolation, and enhanced drive capability, ensuring that the control signal can reliably drive each RIS unit.

[0106] The power transmission module is used to form a focused beam based on a phase control signal to wirelessly power the mobile receiver. The power transmission module includes a radio frequency (RF) source and a RIS array: the RF source generates an RF signal as an energy carrier; the RIS array receives the RF signal, reconstructs the electromagnetic wave front according to the phase control signal issued by the control module, and forms a focused beam pointing towards the predicted position coordinates to wirelessly power the mobile receiver.

[0107] This application describes a rectifier antenna on a mobile vehicle that receives radio frequency energy and converts it into direct current to power loads on the vehicle, such as batteries or sensors.

[0108] The reconfigurable smart surface energy transfer system proposed in this application operates in a fixed cycle period Δt = 100 ms for non-line-of-sight environments. Figure 7 This is a schematic diagram of the FPGA control timing proposed in Embodiment 1 of the present invention. During time slice allocation, approximately 90ms (90%) of each cycle is used for energy transfer, during which the RIS beam remains focused; the remaining 10ms (10%) is used for prediction updates. During this time period, the system receives Bluetooth feedback, runs the prediction algorithm, and updates the RIS phase configuration. The FPGA internally precisely coordinates feedback reception, data preprocessing, prediction calculation (or receiving results from the CPU), 595 data shifting, and latch updates to ensure that all operations are completed within the 10ms update window.

[0109] Embodiment 1 of this invention proposes a reconfigurable intelligent surface energy transfer system for non-line-of-sight environments. Through the synergy of Bluetooth real-time feedback, three-dimensional motion prediction algorithm and FPGA-based large-scale array synchronous control, it achieves low-latency, high-precision energy focusing and tracking of moving targets.

[0110] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that the elements inherent in a process, method, article, or apparatus that includes a list of elements are included. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Additionally, portions of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0111] While specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art can make other modifications or variations based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A reconfigurable smart surface energy transfer system for non-line-of-sight environments, characterized in that, include: The feedback module, deployed at the mobile receiver, collects voltage data of the local load through a voltage sampling circuit and transmits it back to the global scanning module and the three-dimensional prediction module respectively. The global scanning module is used to control the RIS to perform beam scanning and determine the direction of maximum received power based on the voltage data as the initial focusing point; A 3D prediction module, communicatively connected to the feedback module and the global scanning module, establishes a 3D voxel grid centered on the initial focal point. Based on the historical and real-time voltage data sequences continuously transmitted back by the feedback module, it estimates the predicted position coordinates of the mobile receiver using a preset time-series prediction model. The detailed process executed by the 3D prediction module includes: establishing a 3D voxel grid centered on the initial focal point; at each time step, mapping the historical and real-time voltage data sequences from the feedback module onto the 3D voxel grid to form an observation tensor, and extracting spatial features through an encoder to obtain a feature tensor; inputting the feature tensor into the preset time-series prediction model to predict the position probability distribution of the mobile receiver at the predicted time step; and calculating the weighted expectation of the position probability distribution at any time step to obtain the estimated predicted position coordinates. The temporal prediction model in the three-dimensional prediction module is a convolutional long short-term memory network, specifically including: Input Gate: ; Forgotten Gate: ; Output gate: ; Candidate memories: ; Cell status update: ; Hidden status update: ; in, Index for the current time step; express Time input gate, express The Gate of Forgetting Time; express The output gate at any given time; express Candidate memories of a moment; It is the sigmoid activation function; express The convolution kernel weight matrix to the input gate; express The convolution kernel weight matrix leading to the forget gate; express The convolution kernel weight matrix leading to the output gate; express The convolution kernel weight matrix to the candidate state; Indicates the previous hidden state The convolution kernel weight matrix to the input gate; Indicates the previous hidden state The convolution kernel weight matrix leading to the forget gate; Indicates the previous hidden state The convolution kernel weight matrix leading to the output gate; Indicates the previous hidden state The convolution kernel weight matrix to the candidate state; This represents the bias vector of the input gate; The bias vector representing the forget gate; This represents the bias vector of the output gate; The bias vector representing the candidate state; express The time is determined by historical location, received voltage, and RIS status information in a three-dimensional grid. The input feature tensor formed above; express Cellular state at any given moment; Represents the hyperbolic tangent activation function; This represents element-wise multiplication; The multi-objective loss function during the training process of the 3D prediction module is: ； in, Represents cross-entropy loss; This indicates the position mean square error loss; This indicates a loss of speed consistency. Indicates power regression loss; Indicates the L2 regularization term; The weight coefficients representing the cross-entropy loss; The weighting coefficients representing the position mean square error loss; Weighting coefficients representing the speed consistency loss; The weighting coefficients representing the power regression loss; This represents the weight coefficient of the L2 regularization term; The control module, which is communicatively connected to the three-dimensional prediction module, is used to calculate the phase configuration of the RIS unit based on the predicted position coordinates, and synchronously output the phase control signal through a shift register to drive the RIS array. The power transmission module is used to form a focused beam based on the phase control signal to wirelessly power the mobile receiver.

2. The system according to claim 1, characterized in that, The feedback module includes: The ARM microcontroller acquires voltage data from the local load through a voltage sampling circuit. The Bluetooth communication unit, connected to the ARM microcontroller, is used to transmit the collected voltage data back to the global scanning module and the three-dimensional prediction module, respectively.

3. The system according to claim 1, characterized in that, The detailed process executed by the global scanning module includes: The RIS array is controlled to load multiple sets of phase configurations from a pre-stored beamforming codebook in sequence, with each set of phase configurations corresponding to a different beam direction; Iterate through all voltage data received during all phase configuration processes and determine the beam direction corresponding to the maximum voltage value as the initial focus point.

4. The system according to claim 1, characterized in that, The detailed process executed by the control module includes: Based on the predicted location coordinates Calculate the required phase configuration vector for each cell in the RIS cell array. ;in, This indicates the total number of cells contained in the RIS cell array; Represents the first RISC cell array The phase offset that each unit needs to be configured with; Phase configuration vector Each consecutive phase value in Quantized into corresponding digital control words and all Combined into data frames suitable for serial communication; The packaged serial data is shifted to a position driven by a clock signal. In a register chain composed of cascaded shift register chips, once all data is in place, a global latch signal is generated, causing all shift registers to synchronously update the data in their internal registers to the parallel output. The level signal output in parallel from the shift register is passed through a driver circuit and then applied to each corresponding cell in the RIS cell array, so that all cells operate according to their digital control words. Synchronously switch to a new electromagnetic state, thereby forming a directional... The focused beam.

5. The system according to claim 4, characterized in that, The digital control word The calculation process is as follows: ； in, This indicates the bit width of the digital control word that controls a single RIS unit; This function represents rounding down to the nearest integer.

6. The system according to claim 4, characterized in that, The calculation process for the phase offset is as follows: ； in, The electromagnetic carrier wavelength that indicates the system's operation; Indicates the first The total propagation path difference from the reference cell to the target point for each RIS cell; ； Indicates the first The center coordinates of each RIS unit; Indicates the location coordinates of the radio frequency source; Indicates the position coordinates of a pre-selected reference RIS cell; Operators for Euclidean norms.

7. The system according to claim 1, characterized in that, The energy transfer module includes a radio frequency source and a RIS array: The radio frequency source is used to generate radio frequency signals as energy carriers; The RIS array is used to receive the radio frequency signal, reconstruct the electromagnetic wave front according to the phase control signal issued by the control module, form a focused beam pointing to the predicted position coordinates, and provide wireless power to the mobile receiver.

Images