Ray tracing simulation method based on voxelization scene

Abstract

The invention discloses a ray tracing simulation method based on a voxelization scene, and belongs to the technical field of wireless communication, and the method comprises the steps: S1, carrying out the collection and digital reconstruction of channel and three-dimensional environment data through a channel-environment synchronous measurement device, and carrying out the fine processing of a reconstructed digital environment; s2, adding an electromagnetic characteristic matched with an object type to the finely processed scene, and further voxelizing the scene; s3, performing separated efficient ray path search on the voxelization scene with the additional electromagnetic property, and extracting and calibrating the electromagnetic property of the material in the object based on the channel data and path information; and S4, carrying out channel parameter calculation based on the stored simulation path set. According to the method, the problems of complex environment modeling, low calculation efficiency and poor flexibility of the existing ray tracing technology are solved, rapid and automatic modeling and efficient and accurate channel simulation of a real physical scene can be realized, and the authenticity and flexibility of channel modeling are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, specifically to a ray tracing simulation method based on voxelized scenes. Background Technology

[0002] With the development of fifth-generation (5G) and future sixth-generation (6G) mobile communication systems, the application of high-frequency bands such as millimeter waves and terahertz frequencies is becoming increasingly widespread, and antenna scale is evolving towards ultra-large-scale multiple-input multiple-output (UMIM). Against this backdrop, the latency, angle, and power characteristics of wireless channels have become exceptionally complex, playing a decisive role in the performance of communication systems. Therefore, establishing accurate, efficient channel models that reflect the details of specific scenarios is crucial for system design, algorithm verification, and network optimization.

[0003] Currently, channel modeling methods are mainly divided into three categories: empirical / statistical models, deterministic models, and hybrid models. Empirical / statistical models are derived from a large amount of measurement data and are simple to calculate, but they cannot accurately reflect the channel details of a specific physical environment. Deterministic models, especially those based on ray tracing, can provide high-precision, scene-dependent channel characteristics by simulating the physical processes of electromagnetic wave propagation, reflection, and diffraction in the environment, and are considered an important direction for future channel modeling.

[0004] However, traditional ray tracing methods face the following challenges: 1) Complex and time-consuming environmental modeling: Traditional methods typically require accurate 3D CAD models as input. For complex indoor and outdoor scenes that have already been built, manually creating or acquiring such models is costly and time-consuming, making it difficult to achieve rapid deployment and updates; 2) High computational complexity: The intersection calculation between rays and a large number of complex geometric elements (such as triangular facets) is extremely time-consuming, especially in the high-frequency band, where multi-order reflection and diffraction need to be considered, resulting in an exponential increase in computational load; 3) High coupling of the simulation process: Traditional ray tracing simulations typically tightly couple path search with electromagnetic field strength calculation. When it is necessary to fine-tune parameters such as antenna pattern, transmission power, or frequency, even if the propagation path remains unchanged, the entire time-consuming ray tracing process must be rerun from scratch, leading to low optimization iteration efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a ray tracing simulation method based on voxelized scenes, which can realize rapid and automated modeling of real physical scenes and efficient and accurate channel simulation, thus solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A ray tracing simulation method based on voxelized scenes includes:

[0008] S1: Use channel-environment synchronous measurement equipment to acquire and digitally reconstruct channel and three-dimensional environment data, and perform fine processing on the reconstructed digital environment;

[0009] S2: Add electromagnetic properties matching the object type to the finely processed scene and voxelize the scene;

[0010] S3: Perform separate, efficient ray path search on voxelized scenes with additional electromagnetic properties, and extract and calibrate the material electromagnetic properties in objects based on channel data and path information.

[0011] S4: Calculate channel parameters based on the saved simulation path set.

[0012] Preferably, in step S1, a channel-environment synchronization measurement device is used to acquire and digitally reconstruct channel and three-dimensional environment data, and the reconstructed digital environment is then finely processed, including:

[0013] S11: By combining LiDAR scanning equipment with real-time positioning and map building algorithms, the target scene is scanned to obtain three-dimensional environmental point cloud data;

[0014] S12: Filter and denoise the three-dimensional environment point cloud data;

[0015] S13: Perform semantic segmentation on the denoised point cloud data, dividing the point cloud into independent point cloud clusters representing different physical entities, and execute the DBSCN algorithm to cluster the same type of point cloud clusters into independent entities, and initially configure the corresponding standard electromagnetic material properties according to the object type.

[0016] Preferably, in step S2, adding electromagnetic properties matching the object type to the finely processed scene and voxelizing the scene includes:

[0017] S21: Calculate the concave shell boundary of the three-dimensional environment point cloud data on the two-dimensional plane, and determine the three-dimensional voxelization space range in combination with the point cloud height range;

[0018] S22: Generate a three-dimensional voxel mesh within the three-dimensional voxelization space according to the preset voxel size;

[0019] S23: Traverse each voxel in the three-dimensional voxel mesh, determine whether the voxel is occupied based on the number of point clouds contained inside, and assign the electromagnetic material properties of the internal point clouds to the occupied voxels and construct a three-dimensional voxel model.

[0020] Specifically, the scene with material information is voxelized, and irregular point cloud data is converted into regular, discretized 3D voxel models to facilitate the efficient execution of ray tracing algorithms.

[0021] Preferably, in step S3, a separate, efficient ray path search is performed on the voxelized scene with additional electromagnetic properties, and the electromagnetic properties of the material in the object are extracted and calibrated based on channel data and path information, including:

[0022] S31: Set the positions of the transmitter TX and receiver RX in the three-dimensional voxel model, and uniformly emit a large number of initial rays from the transmitter TX position into space;

[0023] S32: The voxel ray tracing algorithm is used to trace each ray in the three-dimensional voxel mesh, and interactive decision-making and processing are performed;

[0024] The interactive decision and processing include: when a ray collides with an occupied voxel, local plane fitting is performed based on the coordinates of the center points of the mutual voxel and multiple occupied voxels in the neighborhood to determine the surface normal vector of the interaction point; the reflection direction is calculated and a new reflected ray is generated based on the incident direction of the ray and the surface normal vector; and a certain number of scattered rays are generated at the interaction point position according to the scattering mode configured in the simulation. When a ray collides with an edge voxel, an additional diffraction path is generated that points directly from the collision point to the receiver RX position.

[0025] S33: Recursively track newly generated rays until the preset maximum number of bounces is reached, the energy is below the threshold, or the ray leaves the scene boundary, and record all valid paths that successfully connect the transmitter TX and the receiver RX.

[0026] S34: Employ multi-threaded parallel computing technology to accelerate the path search process until the ray emitted from all points is tracked and the path calculation results corresponding to each point are saved;

[0027] S35: Based on channel synchronization measurement data and multi-point path search results, continuously track the energy attenuation caused by the same object in different incident and outgoing directions and calculate the reflection coefficient, and inversely calculate the relative permittivity of the material.

[0028] Preferably, in step S4, the channel parameter calculation based on the saved simulation path set includes:

[0029] S41: Read path information from the recorded valid paths;

[0030] S42: For each valid path, calculate the complex gain of the path based on geometric information, the electromagnetic material properties of the interaction points along the path, and the radiation patterns of the transmitting and receiving antennas:

[0031]

[0032] in, Antenna gain; It is the direction of departure of the transmitting antenna along a certain path. Gain on; It is a unit direction vector from the previous propagation node to the next propagation node, where the node can be a transmitter, a receiver, or an interaction point with the environment during the path propagation process; It is the direction of arrival of the receiving antenna along a certain path. The gain can be obtained from a predefined antenna pattern function; It is the free space path loss term. It is the total geometric length of a certain path, that is, the sum of the distances along the path from the transmitter TX to the receiver RX; On the path The cumulative effect of multiple interactions; It is a 2x2 Jones matrix used to describe the polarization and amplitude / phase changes of the electromagnetic wave at each interaction; for reflection events, The elements are composed of Fresnel reflection coefficients and are divided into vertical polarization and parallel polarization, depending on the incident angle. Complex relative permittivity of interacting voxels and conductivity Calculation, where :

[0033]

[0034] For diffraction events, This represents a diffraction coefficient matrix, using the unified diffraction theory diffraction coefficients;

[0035] S43: For the above path information, after calculating the complex gain and propagation delay of all effective propagation paths, the contributions of all paths are coherently superimposed to synthesize the total channel impulse response, which is a time-domain function containing the amplitude, phase, and delay information of all multipath components, expressed as:

[0036]

[0037] in, It is the Dirac impulse function.

[0038] Specifically, channel parameters are calculated based on the path set, which utilizes the stored path set for efficient channel parameter calculation.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] 1. This invention features automated and highly efficient modeling. It uses LiDAR+SLAM technology to replace traditional manual CAD modeling, enabling rapid and automated 3D digitization of any real-world scene. This significantly improves modeling efficiency, increases simulation speed, and separates pathfinding from field calculation. This allows existing path search results to be reused directly when adjusting non-geometric parameters such as antenna parameters and frequency, avoiding repetitive and time-consuming ray tracing. This greatly improves the iterative efficiency of simulation and optimization. At the same time, the voxel-based tracking algorithm and multi-threaded parallel acceleration technology further shorten the computation time.

[0041] 2. The results of this invention are accurate and close to reality. Since the environmental model is directly derived from a precise scan of the real world, and the simulation process takes into account the actual electromagnetic material of the object and the multi-order propagation effect, the obtained channel model has high fidelity and can accurately reflect the channel characteristics of a specific site. It is flexible and scalable, and the voxel resolution can be adjusted according to the requirements of accuracy and efficiency. The method framework is clear and easy to integrate more refined physical models, and has good scalability. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the overall process of the ray tracing simulation method based on voxelized scenes of the present invention;

[0043] Figure 2 This is the channel-environment synchronization detection device used in this invention.

[0044] Figure 3 This is a flowchart of the SLAM algorithm used in this invention for digital environment reconstruction based on real-time radar detection data and IMU data.

[0045] Figure 4 This is a schematic diagram of the environmental point cloud after SLAM mapping and preliminary filtering and noise reduction processing according to the present invention.

[0046] Figure 5 This is a schematic diagram of the environmental point cloud after object segmentation of the point cloud and assignment of corresponding electromagnetic materials according to the object type, as per the present invention.

[0047] Figure 6 A schematic diagram of the point cloud concave shell boundary calculated for the automated determination of the voxel generation range of the present invention.

[0048] Figure 7 This is a schematic diagram of the voxelized environment model generated by converting point clouds with material information according to the present invention.

[0049] Figure 8 This is a schematic diagram illustrating the ray tracing process of the ray tracing simulation method of the present invention.

[0050] Figure 9This is a schematic diagram of the visualization results of the third-order bouncing ray path executed in a voxelized environment according to the present invention;

[0051] Figure 10 This is a schematic diagram of the process for object segmentation and material electromagnetic property extraction based on synchronous channel measured data and environment according to the present invention.

[0052] Figure 11 This is a graph comparing the channel impulse response obtained from the simulation of this invention with the results from professional simulation software and actual measurements. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] To address the issues of complex environment modeling, low computational efficiency, and poor flexibility in existing ray tracing techniques, please refer to... Figures 1-11 This embodiment provides the following technical solution:

[0055] like Figure 1 As shown, a ray tracing simulation method based on voxelized scenes includes:

[0056] S1: Scene Channel - Synchronous Acquisition and Refined Preprocessing of 3D Data;

[0057] This step is the foundation of the entire channel modeling process. It aims to obtain the actual channel data of the scene as a reference and transform the real physical scene into a high-precision digital point cloud model with semantic information.

[0058] First use as follows Figure 2 The synchronous measurement system shown includes lidar and channel detection equipment to collect raw channel and environmental data and uses the SLAM algorithm for data acquisition and environmental reconstruction.

[0059] In this embodiment, a mobile platform integrating a high-precision LiDAR sensor (Ouster OS1-128) and an inertial measurement unit (IMU) is used for measurement. The operator carries the device and performs mobile scanning within the target indoor scene. The LiDAR sensor has 128 scan lines, a pitch angle of -25 to 25 degrees, and a horizontal angle of 360 degrees. It is responsible for emitting laser pulses and receiving reflected light signals. By measuring the time of flight, the distance to the object is accurately calculated, thereby acquiring the three-dimensional coordinate information of a large number of points in the scene, forming a raw point cloud. The IMU provides the device's angular velocity and linear acceleration information to assist in pose estimation. After unpacking the raw data, the usage process is as follows: Figure 3 The SLAM algorithm shown processes LiDAR and IMU data in real time, simultaneously estimating the device's pose and constructing an environmental map. By matching point cloud data from consecutive frames and combining it with IMU data for motion prediction and correction, the SLAM algorithm can overcome sensor noise and accumulated errors, ultimately generating a globally consistent, dense, and accurate 3D environmental point cloud. This point cloud data contains the geometric shape and spatial distribution information of all visible objects in the scene.

[0060] The acquired raw point cloud data may contain outliers or measurement noise that do not belong to the target scene. To improve the efficiency and accuracy of subsequent processing, preliminary filtering is required. Based on the actual physical boundaries of the scene under test, a range of three-dimensional coordinate axes is set. A pass-through filter removes all point cloud data with coordinate values ​​outside this range. This helps remove extreme outliers caused by scanning equipment exceeding scene boundaries or sensor misreading, ensuring that subsequent processed data is focused on the target scene. After pass-through filtering, sparse outliers may still exist in the point cloud due to sensor random noise or environmental factors. A statistical outlier removal algorithm is used to analyze the statistical relationship between each point and its neighbors to identify and remove outliers. The point cloud data becomes smoother and cleaner. See [link to relevant documentation]. Figure 4 It can more accurately reflect the geometric features of the scene.

[0061] The refined point cloud is further processed with semantic segmentation and electromagnetic material attribute assignment. This is a crucial step in transforming geometric information into electromagnetically meaningful data, laying the foundation for subsequent electromagnetic simulations. The DBSCAN clustering algorithm is used to cluster the denoised point cloud data, dividing it into multiple independent clusters. Ideally, each cluster corresponds to an independent physical entity in the scene, such as a wall, a table, a chair, or a window. For each segmented point cloud cluster, its physical entity category is determined based on its geometric features, size, shape, or image recognition. Then, based on these categories, preset electromagnetic material attributes are assigned to all points within the cluster. (See [link to relevant documentation]). Figure 5These properties are essential for electromagnetic wave propagation simulations, including complex relative permittivity and conductivity.

[0062] S2: Construct a voxelized scene with material information using finely processed point cloud data;

[0063] Voxelization transforms irregular point cloud data into a regular, discretized 3D mesh model. This is fundamental to the efficient execution of ray tracing algorithms. To avoid generating unnecessary voxels outside the scene, thus saving storage space and computational resources, the boundaries of the voxel mesh need to be accurately determined. The X and Y coordinates of all preprocessed point clouds with material information are extracted using a 2D concave hull algorithm to form a 2D point set. Then, the concave hull algorithm is used to calculate the concave hull boundary of this 2D point set, and the calculated 2D concave hull boundary is extended to 3D space. See [link to documentation]. Figure 6 By combining the minimum and maximum Z-coordinates of the point cloud, a precise 3D bounding box is defined as the generation region for the voxelized mesh. Users set a global voxel size based on their simulation accuracy and computational efficiency requirements. Voxel size is a key parameter affecting simulation accuracy and computational cost; smaller voxels retain more scene details, but the number of generated voxels increases cubically, leading to increased computational and storage demands. Based on the defined 3D spatial range and the set voxel size, this space is divided into a uniform 3D voxel mesh. Each voxel is a cube unit with a side length equal to the voxel size. See [link to documentation]. Figure 7 Each generated voxel unit is traversed to determine whether it is occupied by a physical entity in the scene and inherits the corresponding electromagnetic properties. For each voxel, the number of point clouds falling into it is counted. If the number of point clouds contained in a voxel exceeds a preset density threshold (e.g., exceeding 1% of the total number of points), the voxel is marked as "occupied", indicating that there is a physical entity inside the voxel. Otherwise, it is marked as "idle". For voxels marked as "occupied", electromagnetic material properties need to be assigned to them. For each occupied voxel, one or more point cloud points whose center point is closest to it in the KD tree are found. Then, the electromagnetic material properties carried by these point cloud points are assigned to the voxel.

[0064] S3: Perform efficient ray path search based on voxelized scenes;

[0065] This step is the core of the invention, aiming to efficiently find all valid propagation paths from transmitter TX to receiver RX, and supporting multi-stage bouncing and parallel acceleration. The three-dimensional spatial coordinates of transmitter TX and receiver RX are precisely set in the voxelized scene. Transmitter TX and receiver RX can be represented as point sources or regions of specific sizes. A large number of initial rays are uniformly emitted from the transmitter TX position into the three-dimensional space. To ensure the comprehensiveness of the path search, the number of initial rays is usually very large. The emission direction is uniformly sampled using the Fibonacci method, and rays with equal angular intervals are emitted. Based on a digital differential analyzer, the next voxel the ray passes through is efficiently determined, avoiding complex floating-point calculations. The basic idea is to calculate the intersection of the ray and the voxel grid line for a given ray, and to advance the ray in the grid with a fixed step size. To further accelerate collision detection, a multi-resolution hash grid is pre-constructed in the space, mapping the coordinates of voxels to a hash table. This allows the ray to quickly query candidate voxels that may collide with a complexity close to O(1) during its movement, without having to traverse all voxels in the entire scene. When the ray moves in the voxel grid, it needs to determine whether it interacts with an occupied voxel and perform corresponding processing based on the interaction type. When the ray enters an occupied voxel from an empty voxel, or when the ray moves inside an occupied voxel, it is determined whether the ray path is close enough to the center of the occupied voxel. Once a collision is detected, an interaction event occurs.

[0066] For determining the normals of interactive voxels, a more robust method is adopted: Using the voxel where the interaction point is located as the center, the coordinates of the center points of multiple occupied voxels in its surrounding neighborhood are selected. Using these center points, a local plane fitting algorithm is used to estimate the local surface normal vector at the interaction point. This method can better handle surface roughness or non-planar structures in the original scene that may exist in the voxelized model, improving the accuracy of the reflection direction. After interaction, a normally symmetrical reflected ray and a scattered ray conforming to the set scattering mode are generated at the center of the interactive voxel, and the ray continues to be tracked. The newly generated reflected and diffracted rays continue the above tracking and interaction process, thereby simulating multi-order reflection and diffraction effects. This process is recursive until the preset maximum number of bounces is reached, the ray energy is below a threshold, or the ray leaves the scene boundary. See [link to relevant documentation]. Figure 8 A valid propagation path is considered found when any ray eventually comes close enough to the receiver RX. This path is completely recorded and stored in a path database. The record information for each path includes: the coordinates of the path's starting point (transmitter TX) and ending point (receiver RX), a series of ordered interaction point coordinates, the type of each interaction point (backscattering, diffraction), the normal vector at each interaction point, the material properties of the voxels along the path, the total geometric length of the path, and the initial polarization information of the path. For visualization results after ray tracing, see [link to visualization]. Figure 9 To significantly shorten pathfinding time, multi-threading technology is used to parallelize the ray tracing process, making full use of the computing power of modern multi-core processors. The millions of rays initially emitted can be evenly grouped, and different ray tracing tasks are assigned to multiple CPU cores of the computer for parallel processing. Each thread or process independently traces the ray it is assigned to, and each thread or process stores the effective path it finds in a shared data structure, or merges the results after each thread or process completes its task. This parallelization strategy can reduce the total pathfinding time by several times or even tens of times, enabling the complex simulation of large-scale scenes and high-order bounces to be completed within an acceptable time.

[0067] After obtaining path interaction information, it is necessary to further obtain the electromagnetic properties of various objects in the environment to achieve a digital twin of the radio electromagnetic environment. However, traditional material electromagnetic parameters usually require measurement in an anechoic chamber or in-situ measurement of specific reflective structures to assess reflection loss, which is difficult to apply to efficient electromagnetic parameter estimation and digital twin of actual three-dimensional environments. Therefore, this invention uses a fast electromagnetic parameter estimation method based on deterministic computation and joint analysis of measured data. By performing necessary limited channel measurements in the environment to obtain multipath information, and solving the retroreflection path of the first-order reflection to estimate the electromagnetic parameters of the corresponding reflection position, the accurate three-dimensional environment obtained in previous studies is used, combined with deterministic computation, to achieve automatic spatial mapping of multipath and reflectors in the environment and electromagnetic parameter matching and assignment. The above processing is as follows: Figure 10 As shown.

[0068] S4: Channel parameter calculation based on path set;

[0069] This step is another core element of the simulation framework of this invention. It is executed independently of the previous step, utilizing the stored path set for efficient channel parameter calculation. All valid propagation paths searched in S3 are stored in a persistent path database. This reflects the advantage of the invention's "separation of path finding and field calculation." When the user needs to adjust geometry-independent parameters (such as operating frequency, transmit power, or the electromagnetic material properties of certain objects in the scene), there is no need to rerun the time-consuming ray tracing process. Instead, the existing path geometry information is read from the path database, and then the field calculation is re-performed based on the new parameters. This greatly improves the efficiency and flexibility of simulation iteration. For any path in the path database, its total complex gain is a complex value, including the path's amplitude attenuation and phase delay. Finally, the complex gains of all searched paths are coherently superimposed to synthesize the total channel impulse response.

[0070] The total complex gain for any path in the path database can be expressed as:

[0071]

[0072] in, Antenna gain; It is the direction of departure of the transmitting antenna along a certain path. Gain on; It is a unit direction vector from the previous propagation node to the next propagation node, where the node can be a transmitter, a receiver, or an interaction point with the environment during the path propagation process; It is the direction of arrival of the receiving antenna along a certain path. The gain can be obtained from a predefined antenna pattern function; It is the free space path loss term. It is the total geometric length of a certain path, that is, the sum of the distances along the path from the transmitter TX to the receiver RX; On the path The cumulative effect of multiple interactions; It is a 2x2 Jones matrix used to describe the polarization and amplitude / phase changes of the electromagnetic wave at each interaction; for reflection events, The elements are composed of Fresnel reflection coefficients and are divided into vertical polarization and parallel polarization, depending on the incident angle. Complex relative permittivity of interacting voxels and conductivity Calculation, where :

[0073]

[0074] For diffraction events, This represents a diffraction coefficient matrix; the present invention uses the diffraction coefficients of the unified diffraction theory.

[0075] For the aforementioned path information, after calculating the complex gain and propagation delay of all effective propagation paths, the contributions of all paths are coherently superimposed to synthesize the total channel impulse response. The channel impulse response is a time-domain function representing the channel's response to an ideal impulse signal, containing the amplitude, phase, and delay information of all multipath components, and can be expressed as:

[0076]

[0077] in, It is the Dirac impulse function, obtained by... By performing a Fourier transform, the channel frequency response can be obtained, which can be used for broadband system analysis and channel modeling such as OFDM.

[0078] To verify the accuracy and reliability of the method of this invention, the simulated channel impulse response (CIR) was compared with the results of professional ray tracing simulation software and actual field measurements. (See [link to relevant documentation]). Figure 11 The CIR obtained by the method of the present invention shows good consistency with commercial software and actual measurement results in terms of delay, amplitude and overall multipath structure of the main path. This proves the effectiveness and high accuracy of the present invention in predicting wireless channel characteristics.

[0079] In summary, this invention successfully realizes a fast, accurate, and flexible wireless channel modeling method by combining real-world 3D perception technology (LiDAR+SLAM), efficient voxelized scene representation, an innovative pathfinding and field computation separation framework, and an optimized ray tracing algorithm. This method not only solves the bottlenecks of traditional ray tracing in environmental modeling and computational efficiency, but also provides a powerful tool for the design, optimization, and deployment of future wireless communication systems such as 5G / 6G by simulating the electromagnetic wave propagation process with high fidelity. It has significant value in both theoretical research and engineering applications.

[0080] It should be noted that, in this document, relational terms such as "first" and "second" are used only 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 a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0081] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A ray tracing simulation method based on voxelized scenes, characterized in that, include: S1: Use channel-environment synchronous measurement equipment to acquire and digitally reconstruct channel and three-dimensional environment data, and perform fine processing on the reconstructed digital environment; S2: Add electromagnetic properties matching the object type to the finely processed scene and voxelize the scene; S3: Perform separate, efficient ray path search on voxelized scenes with additional electromagnetic properties, and extract and calibrate the material electromagnetic properties in objects based on channel data and path information. S4: Calculate channel parameters based on the saved simulation path set.

2. The ray tracing simulation method based on voxelized scenes according to claim 1, characterized in that, In step S1, a channel-environment synchronization measurement device is used to acquire and digitally reconstruct channel and three-dimensional environment data, and the reconstructed digital environment is then finely processed, including: S11: By combining LiDAR scanning equipment with real-time positioning and map building algorithms, the target scene is scanned to obtain three-dimensional environmental point cloud data; S12: Filter and denoise the three-dimensional environment point cloud data; S13: Perform semantic segmentation on the denoised point cloud data, dividing the point cloud into independent point cloud clusters representing different physical entities, and execute the DBSCN algorithm to cluster the same type of point cloud clusters into independent entities, and initially configure the corresponding standard electromagnetic material properties according to the object type.

3. The ray tracing simulation method based on voxelized scenes according to claim 2, characterized in that, In step S2, electromagnetic properties matching the object types are added to the finely processed scene, and the scene is voxelized, including: S21: Calculate the concave shell boundary of the three-dimensional environment point cloud data on the two-dimensional plane, and determine the three-dimensional voxelization space range in combination with the point cloud height range; S22: Generate a three-dimensional voxel mesh within the three-dimensional voxelization space according to the preset voxel size; S23: Traverse each voxel in the three-dimensional voxel mesh, determine whether the voxel is occupied based on the number of point clouds contained inside, and assign the electromagnetic material properties of the internal point clouds to the occupied voxels and construct a three-dimensional voxel model.

4. The ray tracing simulation method based on voxelized scenes according to claim 3, characterized in that, In step S3, a separate, efficient ray path search is performed on the voxelized scene with additional electromagnetic properties, and the electromagnetic properties of the material in the object are extracted and calibrated based on channel data and path information, including: S31: Set the positions of the transmitter TX and receiver RX in the three-dimensional voxel model, and uniformly emit a large number of initial rays from the transmitter TX position into space; S32: The voxel ray tracing algorithm is used to trace each ray in the three-dimensional voxel mesh, and interactive decision-making and processing are performed; The interactive decision and processing include: when a ray collides with an occupied voxel, local plane fitting is performed based on the coordinates of the center points of the mutual voxel and multiple occupied voxels in the neighborhood to determine the surface normal vector of the interaction point; the reflection direction is calculated and a new reflected ray is generated based on the incident direction of the ray and the surface normal vector; and a certain number of scattered rays are generated at the interaction point position according to the scattering mode configured in the simulation. When a ray collides with an edge voxel, an additional diffraction path is generated that points directly from the collision point to the receiver RX position. S33: Recursively track newly generated rays until the preset maximum number of bounces is reached, the energy is below the threshold, or the ray leaves the scene boundary, and record all valid paths that successfully connect the transmitter TX and the receiver RX. S34: Employ multi-threaded parallel computing technology to accelerate the path search process until the ray emitted from all points is tracked and the path calculation results corresponding to each point are saved; S35: Based on channel synchronization measurement data and multi-point path search results, continuously track the energy attenuation caused by the same object in different incident and outgoing directions and calculate the reflection coefficient, and inversely calculate the relative permittivity of the material.

5. The ray tracing simulation method based on voxelized scenes according to claim 4, characterized in that, In step S4, channel parameters are calculated based on the saved simulation path set, including: S41: Read path information from the recorded valid paths; S42: For each valid path, calculate the complex gain of the path based on geometric information, the electromagnetic material properties of the interaction points along the path, and the radiation patterns of the transmitting and receiving antennas: ； in, Antenna gain; It is the direction of departure of the transmitting antenna along a certain path. Gain on; It is a unit direction vector from the previous propagation node to the next propagation node, where the node can be a transmitter, a receiver, or an interaction point with the environment during the path propagation process; It is the direction of arrival of the receiving antenna along a certain path. The gain can be obtained from a predefined antenna pattern function; It is the free space path loss term. It is the total geometric length of a certain path, that is, the sum of the distances along the path from the transmitter TX to the receiver RX; On the path The cumulative effect of multiple interactions; It is a 2x2 Jones matrix used to describe the polarization and amplitude / phase changes of the electromagnetic wave at each interaction; for reflection events, The elements are composed of Fresnel reflection coefficients and are divided into vertical polarization and parallel polarization, depending on the incident angle. Complex relative permittivity of interacting voxels and conductivity Calculation, where : ； For diffraction events, This represents a diffraction coefficient matrix, using the unified diffraction theory diffraction coefficients; S43: For the above path information, after calculating the complex gain and propagation delay of all effective propagation paths, the contributions of all paths are coherently superimposed to synthesize the total channel impulse response, which is a time-domain function containing the amplitude, phase, and delay information of all multipath components, expressed as: ； in, It is the Dirac impulse function.

Images