A method, system, electronic device, and storage medium for updating scene state.

Abstract

This application discloses a scene state update method, system, electronic device, and storage medium, belonging to the technical field of computer simulation technology. The scene state update method includes: dividing all charged particles in a particle grid into an arbitrary number of particle clusters, and determining the initial spatial matrix and initial velocity matrix corresponding to each particle cluster; performing singular value decomposition on the initial spatial matrix, and truncating the singular value decomposition result to obtain a first low-rank approximation matrix; performing singular value decomposition on the initial velocity matrix, and truncating the singular value decomposition result to obtain a second low-rank approximation matrix; updating the particle information of the particle grid using the first and second low-rank approximation matrices; and updating the scene state of the simulated scene at the current moment based on the current particle information. This application can reduce the computational load of the particle information update process and improve the computational efficiency of the electronic device.

Description

Technical Field

[0001] This application relates to the field of computer simulation technology, and in particular to a method, system, electronic device and storage medium for updating scene state. Background Technology

[0002] The Particle-in-Cell (PIC) method is a widely used numerical simulation technique in computational physics, plasma physics, and fluid dynamics, primarily used to study the interaction between charged particles and electromagnetic fields. By combining particle simulation and grid computation, this method can effectively describe complex plasma behaviors (such as the motion and interaction of charged particles in electromagnetic fields), thereby determining the state of the simulated scene at various time points and accurately demonstrating the evolution of the scene's state.

[0003] When using particle meshes for scene simulation, obtaining high-resolution output often requires a large number of computational operations, resulting in low computational efficiency.

[0004] Therefore, how to reduce the computational load of the particle information update process and improve the computational efficiency of electronic devices is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a scene state update method, system, electronic device, and storage medium that can reduce the computational load of the particle information update process and improve the computational efficiency of the electronic device.

[0006] To address the aforementioned technical problems, this application provides a scene state update method, applied to an electronic device with singular value decomposition functionality, comprising:

[0007] Obtain scene parameters of the scene to be simulated, and generate a corresponding particle mesh based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh is set with an arbitrary number of charged particles;

[0008] Divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster.

[0009] The initial spatial matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial spatial matrix is ​​truncated to obtain a first low-rank approximate matrix.

[0010] The initial velocity matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial velocity matrix is ​​truncated to obtain a second low-rank approximate matrix.

[0011] The particle information of the particle grid is updated using the first low-rank approximation matrix and the second low-rank approximation matrix;

[0012] The scene state of the simulated scene at the current moment is updated based on the current particle information.

[0013] Optionally, updating the particle information of the particle mesh using the first low-rank approximation matrix and the second low-rank approximation matrix includes:

[0014] The first low-rank approximation matrix is ​​converted into a particle matrix, and the grid point charge density is calculated based on the particle matrix.

[0015] The electric field at the grid points is obtained by solving the Poisson equation using the charge density at the grid points.

[0016] The electric field at the grid point is mapped from the grid point to the position of the charged particle to obtain the electric field experienced by the charged particle;

[0017] The second low-rank approximation matrix is ​​updated using the electric field experienced by the charged particle;

[0018] The first low-rank approximation matrix is ​​updated using the updated second low-rank approximation matrix;

[0019] The particle information of the particle grid is updated based on the updated first low-rank approximation matrix and the updated second low-rank approximation matrix.

[0020] Optionally, the particle information of the particle mesh is updated based on the updated first low-rank approximation matrix and the updated second low-rank approximation matrix, including:

[0021] The current position information of the charged particle is determined based on the updated first low-rank approximation matrix;

[0022] The current velocity information of the charged particle is determined based on the updated second low-rank approximation matrix;

[0023] The particle information of the particle mesh is updated based on the current position and velocity information.

[0024] Optionally, after determining the current position information of the charged particle based on the updated first low-rank approximation matrix, the method further includes:

[0025] Determine whether the current position information of the charged particle is within the particle's motion range;

[0026] If not, the charged particles are remapped into the particle motion range using periodic boundary conditions.

[0027] Optionally, all the charged particles in the particle grid are divided into an arbitrary number of particle clusters, including:

[0028] All charged particles in the particle grid are divided into a particle cluster based on the global domain.

[0029] Alternatively, all the charged particles in the particle grid can be divided into multiple particle clusters according to subdomains;

[0030] Accordingly, the initial spatial matrix and initial velocity matrix corresponding to each particle cluster are determined, including:

[0031] Determine the number of particles in the particle cluster;

[0032] The initial space matrix and the initial velocity matrix are generated for each particle cluster based on the number of particles.

[0033] Optionally, the singular value decomposition result of the initial space matrix is ​​truncated, including:

[0034] A first rank value is determined based on the number of rows and columns of the initial space matrix, and the singular value decomposition result of the initial space matrix is ​​truncated according to the first rank value; wherein, the first rank value is less than the number of rows of the initial space matrix and the first rank value is less than the number of columns of the initial space matrix;

[0035] Accordingly, the singular value decomposition results of the initial velocity matrix are truncated, including:

[0036] The second rank value is determined based on the number of rows and columns of the initial velocity matrix, and the singular value decomposition result of the initial velocity matrix is ​​truncated according to the second rank value; wherein the second rank value is less than the number of rows of the initial velocity matrix and the second rank value is less than the number of columns of the initial velocity matrix.

[0037] This application also provides a scene state update method based on laser plasma instability distribution simulation, applicable to electronic devices with singular value decomposition capabilities, including:

[0038] Obtain scene parameters for simulating laser plasma instability distribution, and generate corresponding particle meshes based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh contains an arbitrary number of charged particles;

[0039] Divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster.

[0040] The initial spatial matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial spatial matrix is ​​truncated to obtain a first low-rank approximate matrix.

[0041] The initial velocity matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial velocity matrix is ​​truncated to obtain a second low-rank approximate matrix.

[0042] The particle information of the particle grid is updated using the first low-rank approximation matrix and the second low-rank approximation matrix;

[0043] The simulation of the laser plasma instability distribution is updated based on the current particle information to reflect the current state of the scenario at the current moment.

[0044] This application also provides a scene state update system, applied to an electronic device with singular value decomposition functionality, including:

[0045] The particle mesh generation module is used to acquire scene parameters of the scene to be simulated and generate a corresponding particle mesh based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh is set with an arbitrary number of charged particles;

[0046] The matrix generation module is used to divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and to determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster.

[0047] The first processing module is used to perform singular value decomposition on the initial space matrix and truncate the singular value decomposition result of the initial space matrix to obtain a first low-rank approximate matrix.

[0048] The second processing module is used to perform singular value decomposition on the initial velocity matrix and truncate the singular value decomposition result of the initial velocity matrix to obtain a second low-rank approximate matrix.

[0049] The particle information update module is used to update the particle information of the particle grid using the first low-rank approximation matrix and the second low-rank approximation matrix;

[0050] The scene state update module is used to update the scene state of the scene to be simulated at the current moment based on the current particle information.

[0051] This application also provides a storage medium storing a computer program thereon, which, when executed, implements the steps of the above-described scenario state update method.

[0052] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor invokes the computer program in the memory to implement the steps of the above-described scenario state update method.

[0053] This application discloses a scene state update method. After generating a particle mesh, the method divides all charged particles in the particle mesh into an arbitrary number of particle clusters and determines the initial spatial matrix and initial velocity matrix corresponding to each particle cluster. This application performs singular value decomposition and truncation operations on the initial spatial matrix and initial velocity matrix to obtain a first low-rank approximation matrix and a second low-rank approximation matrix. The truncation operation on the singular value decomposition results reduces the data dimensionality while retaining key information. The computation and storage cost of the low-rank approximation matrix is ​​much lower than that of the original high-dimensional matrix. The first and second low-rank approximation matrices can be used to quickly update the particle information of the particle mesh, thereby determining the scene state at the current moment. Therefore, this application can reduce the computational load of the particle information update process and improve the computational efficiency of electronic devices. This application also provides a scene state update system, a storage medium, and an electronic device, which have the above-mentioned beneficial effects, and will not be elaborated further here. Attached Figure Description

[0054] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a scene state update method provided in an embodiment of this application;

[0056] Figure 2 This is a flowchart of a low-rank approximation simulation method based on particle mesh, provided as an embodiment of this application. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] The particle mesh method is used to simulate the dynamic behavior of charged particles in plasmas. Traditional fluid dynamics models (such as the Euler equations or Navier-Stokes equations) cannot adequately describe the behavior of charged particles, especially the interactions and electromagnetic effects between them. The particle mesh method solves this problem by introducing the coupling between particles and the electromagnetic field. The core idea of ​​the particle mesh method is to discretize a spatial grid in the computational domain and calculate the electric field on these grids. Then, the charged particles are propelled according to their motion in the electric field, and finally, the electric field is updated. In this way, the particle mesh method can simulate the complex interactions between charged particles and the electromagnetic field on a large scale.

[0059] However, the particle mesh method involves a huge computational load, severely impacting the efficiency of physical simulations. To address the technical problems of the aforementioned particle mesh method, this embodiment provides a particle information update scheme for particle meshes. Please refer to the following... Figure 1 , Figure 1 This is a flowchart of a scene state update method provided in an embodiment of this application.

[0060] Specific steps may include:

[0061] S101: Obtain the scene parameters of the scene to be simulated, and generate the corresponding particle mesh based on the scene parameters;

[0062] This embodiment can be applied to electronic devices with singular value decomposition capabilities to simulate the dynamic behavior of charged particles in the simulated scenario.

[0063] The scenarios to be simulated can include laser-plasma interaction, spacecraft propulsion, semiconductor design, astrophysical motion, and other similar scenarios. The scenario parameters can include information such as particle motion range, particle number, particle velocity, particle position, and particle charge. Based on these parameters, a corresponding particle mesh can be generated and initialized. The particle mesh can contain multiple meshes, and each mesh contains an arbitrary number of charged particles.

[0064] Specifically, in this embodiment, the physical space corresponding to the scene to be simulated can be determined, the physical space can be divided into multiple uniform grids, and an arbitrary number of charged particles can be generated in each grid to generate the particle grid. The grids in the particle grid have the same size and shape.

[0065] S102: Divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and determine the initial space matrix and initial velocity matrix corresponding to each particle cluster.

[0066] This step involves dividing all charged particles in the particle mesh into one or more particle clusters. After initializing the particle mesh and charged particles, this embodiment can determine the initial velocity and initial position of each charged particle in the particle mesh, and summarize the initial positions of all particles in each cluster into a spatial matrix, and summarize the initial velocities of all particles in each cluster into a velocity matrix.

[0067] S103: Perform singular value decomposition on the initial space matrix and truncate the singular value decomposition result of the initial space matrix to obtain a first low-rank approximate matrix;

[0068] After obtaining the initial spatial matrix, this step can perform singular value decomposition on the initial spatial matrix to obtain the singular value decomposition result, and select the k largest singular values ​​and their corresponding singular vectors from the singular value decomposition result, truncate the rest, and construct the first low-rank approximation matrix.

[0069] S104: Perform singular value decomposition on the initial velocity matrix and truncate the singular value decomposition result of the initial velocity matrix to obtain a second low-rank approximation matrix;

[0070] In this step, after obtaining the initial velocity matrix, singular value decomposition is performed on the initial velocity matrix to obtain the singular value decomposition result. The j largest singular values ​​and their corresponding singular vectors are selected from the singular value decomposition result, the rest are truncated, and a second low-rank approximation matrix is ​​constructed.

[0071] S105: Update the particle information of the particle grid using the first low-rank approximation matrix and the second low-rank approximation matrix.

[0072] Based on the first and second low-rank approximation matrices, the grid point charge density, grid point electric field, and electric field experienced by charged particles can be calculated using the first and second low-rank approximation matrices. The first and second low-rank approximation matrices for the next step are then calculated based on the electric field experienced by the charged particles, so as to update the particle information of the particle grid.

[0073] The particle information of the particle mesh includes the position and velocity of the charged particles. After determining the first low-rank approximation matrix and the second low-rank approximation matrix for the next step, the position of the charged particles can be updated according to the latest first low-rank approximation matrix, and the velocity of the charged particles can be updated according to the latest second low-rank approximation matrix.

[0074] S106: Update the scene state of the scene to be simulated at the current moment based on the current particle information.

[0075] After obtaining the current particle information, this embodiment can update the scene state of the simulated scenario at the current moment based on the current particle information. The scene state is used to describe the overall state of all particles at a certain moment during the evolution of the simulated scenario. For example, the scene state may include: the average flow velocity of the fluid, the electromagnetic field strength, the particle number density, the energy spectral density, etc.

[0076] In this embodiment, after generating the particle mesh, all charged particles in the mesh are divided into an arbitrary number of particle clusters, and the initial spatial matrix and initial velocity matrix corresponding to each particle cluster are determined. This embodiment performs singular value decomposition and truncation operations on the initial spatial matrix and initial velocity matrix to obtain a first low-rank approximation matrix and a second low-rank approximation matrix. The truncation operation on the singular value decomposition results reduces the data dimensionality while retaining key information. The computation and storage cost of the low-rank approximation matrix is ​​much lower than that of the original high-dimensional matrix. The first and second low-rank approximation matrices can be used to quickly update the particle information of the particle mesh, thereby determining the scene state at the current moment. Therefore, this embodiment can reduce the computational load of the particle information update process and improve the computational efficiency of electronic devices.

[0077] As for Figure 1 A further description of the corresponding embodiment: the process of dividing all the charged particles in the particle grid into an arbitrary number of particle clusters includes: dividing all the charged particles in the particle grid into one particle cluster according to the whole domain; or, dividing all the charged particles in the particle grid into multiple particle clusters according to subdomains.

[0078] Accordingly, when determining the initial space matrix and initial velocity matrix corresponding to each particle cluster, this embodiment can determine the number of particles in the particle cluster and generate the initial space matrix and initial velocity matrix corresponding to each particle cluster based on the number of particles.

[0079] As for Figure 1 A further description of the corresponding embodiment: the process of truncating the singular value decomposition result of the initial space matrix includes: determining a first rank value based on the number of rows and columns of the initial space matrix, and truncating the singular value decomposition result of the initial space matrix according to the first rank value; wherein, the first rank value is less than the number of rows of the initial space matrix, and the first rank value is less than the number of columns of the initial space matrix.

[0080] Accordingly, the process of truncating the singular value decomposition result of the initial velocity matrix includes: determining a second rank value based on the number of rows and columns of the initial velocity matrix, and truncating the singular value decomposition result of the initial velocity matrix according to the second rank value; wherein, the second rank value is less than the number of rows of the initial velocity matrix, and the second rank value is less than the number of columns of the initial velocity matrix.

[0081] As for Figure 1 A further description of the corresponding embodiment: the process of updating the particle information of the particle mesh using the first low-rank approximation matrix and the second low-rank approximation matrix includes the following steps:

[0082] Step A1: Convert the first low-rank approximation matrix into a particle matrix, and calculate the grid point charge density based on the particle matrix.

[0083] Specifically, this step involves calculating the interpolation function for the particle matrix to obtain the grid point charge density. The grid point charge density refers to the charge density at each grid point, where a grid point is the intersection of grid lines.

[0084] Step A2: Solve the Poisson equation using the grid point charge density to obtain the grid point electric field.

[0085] Step A3: Map the electric field at the grid points to the position of the charged particle to obtain the electric field experienced by the charged particle.

[0086] Specifically, this step uses an interpolation function to map the electric field at the grid points to the particle positions, thus obtaining the electric field experienced by the charged particle.

[0087] Step A4: Update the second low-rank approximation matrix using the electric field experienced by the charged particle.

[0088] Specifically, the velocity and position of charged particles in the particle grid are updated iteratively. Based on the electric field experienced by the charged particles, this embodiment can update the second low-rank approximation matrix according to the electric field experienced by the charged particles at the nth iteration number based on the dynamic low-rank method. The updated second low-rank approximation matrix is ​​used to describe the velocity information of the charged particles at the n+1th iteration number.

[0089] Step A5: Update the first low-rank approximation matrix using the updated second low-rank approximation matrix.

[0090] Specifically, this step can use the basis-addition method to update the first low-rank approximation matrix using the updated second low-rank approximation matrix. The updated first low-rank approximation matrix is ​​used to describe the position information of the charged particles when the iteration number is n+1.

[0091] Furthermore, after determining the current position information of the charged particle based on the updated first low-rank approximation matrix, it can also be determined whether the current position information of the charged particle is within the particle's motion range; if yes, no processing is performed; if no, the charged particle is remapped into the particle's motion range using periodic boundary conditions.

[0092] Step A6: Update the particle information of the particle grid according to the updated first low-rank approximation matrix and the updated second low-rank approximation matrix.

[0093] Specifically, this step can determine the current position information of the charged particle based on the updated first low-rank approximation matrix, determine the current velocity information of the charged particle based on the updated second low-rank approximation matrix, and update the particle information of the particle grid based on the current position information and the current velocity information.

[0094] After executing step A6, it can be determined whether the number of iterations meets the requirements. If not, the operations of steps A1-A6 are executed again.

[0095] This application also provides a scene state update method based on laser plasma instability distribution simulation, applied to electronic devices with singular value decomposition capabilities. This method is for... Figure 1 The specific implementation process of this method in the simulation of laser plasma instability distribution, corresponding to the above embodiments, is the same as that of the embodiments described above. The implementation process of this method is as follows:

[0096] Scene parameters for simulating laser plasma instability distribution are obtained, and a corresponding particle mesh is generated based on these parameters. The scene parameters include particle motion range, particle number, particle velocity, and particle position. Each grid cell contains an arbitrary number of charged particles. All charged particles in the particle mesh are divided into an arbitrary number of particle clusters, and an initial spatial matrix and an initial velocity matrix are determined for each cluster. Singular value decomposition (SVD) is performed on the initial spatial matrix, and the SVD result is truncated to obtain a first low-rank approximation matrix. Singular value decomposition (SVD) is performed on the initial velocity matrix, and the SVD result is truncated to obtain a second low-rank approximation matrix. The particle information of the particle mesh is updated using the first and second low-rank approximation matrices. The scene state of the laser plasma instability distribution simulation at the current moment is updated based on the current particle information.

[0097] The scene parameters obtained in the above embodiments are those for the scenario of simulating laser-plasma instability distribution. The scene state determined in the above embodiments is the state of the laser-plasma instability distribution simulation process at the current moment, such as electron density, regional average temperature, particle velocity distribution, etc.

[0098] The process described in the above embodiments is illustrated below through examples in practical applications.

[0099] The particle mesh method treats fluids as being composed of a large number of discrete particles that move and interact on a spatial mesh, thereby coupling the motion of the particles with the electric or magnetic field of the mesh. This method can effectively simulate the macroscopic behavior of matter and the interaction of microscopic particles.

[0100] In the particle mesh method, particles represent discrete units of matter in a fluid. Each particle carries information such as mass, charge, and momentum, and moves dynamically in space. The mesh describes physical quantities such as electric and magnetic fields. It solves for the values ​​of the electric and magnetic fields at the mesh nodes numerically, then applies this field information to the particles to calculate their electromagnetic forces or other external forces. Particles move within the mesh and adjust their states according to changes in the fields, then the electromagnetic fields are recalculated, completing an iterative process. This alternating update mechanism of particles and mesh enables the particle mesh method to provide high accuracy and flexibility in handling complex fluid-electromagnetic coupling problems, making it particularly suitable for large-scale plasma flows, particle dynamics in particle accelerators, and problems in high-energy physics experiments. With the continuous improvement of computing power, the particle mesh method has become a powerful tool for studying complex physical phenomena.

[0101] Low-rank approximation methods have gained widespread attention in the numerical solution of Vlasov equations and other physical equations. These methods aim to simplify computation by reducing the dimensionality of the system, effectively reducing the required storage and computational resources. When dealing with high-dimensional distribution functions, low-rank approximations can significantly reduce data complexity and storage requirements by finding low-rank approximations, thereby improving the efficiency of numerical computation. The advantage of low-rank approximation methods lies in their ability to preserve important physical characteristics while removing redundant information. In solving the Vlasov-Poisson equations, low-rank approximation methods can effectively capture the dynamic changes of the particle distribution function without having to perform full calculations in high-dimensional space, which is particularly important when dealing with complex plasma behavior. In this way, computational accuracy can be maintained while significantly reducing computation time and resource consumption. To address the high memory and computational demands of the PIC method due to its high particle count, this invention employs a low-rank approximation method based on Singular Value Decomposition (SVD) to compress all particles. Furthermore, it uses dynamic low-rank and basis-addition methods as new equations of motion to evolve over time, thereby accelerating the operation of the PIC method.

[0102] The implementation process of the electrostatic model using the conventional particle mesh method may include the following steps:

[0103] Step A1: Initialize the mesh and particles.

[0104] Specifically, this step includes the process of mesh generation, setting particle distribution, and setting the number of iterations.

[0105] The mesh generation process is as follows: The physical region to be calculated is divided into multiple meshes, each representing a small region of space. Uniform or non-uniform meshes can be used. The mesh size determines the resolution of the interaction between particles and the electromagnetic field; more meshes result in higher resolution, and vice versa.

[0106] The process of setting the particle distribution is as follows: At the start of the simulation, each grid cell is initialized with a set of charged particles. The positions and velocities of these particles are initialized based on the physical scenario of the problem, typically satisfying a Maxwell-Boltzmann distribution or other distributions.

[0107] In this embodiment, the number of iterations n can be set. Executing steps A2 to A5 constitutes one iteration, and a total of n iterations are required.

[0108] Step A2: Calculate the grid point charge density.

[0109] Specifically, in the particle mesh method, particles are discrete, while the electric field is defined in continuous space. Therefore, at each time step, an interpolation operation is required to map the particle's charge from its position to the mesh node. This process is accomplished using common interpolation methods (such as linear interpolation, bilinear interpolation, or higher-order interpolation methods).

[0110] The formula for calculating the point charge density of the grid is as follows:

[0111] ;

[0112] in, For grid points Charge density at that location, Here, N is the charge of the i-th particle, N is the total number of particles, and S is the interpolation method used to calculate the charge of the i-th particle. To grid point The weighting coefficient at the location.

[0113] Step A3: Calculate the electric field at the grid points (i.e., the spatial electric field).

[0114] Within each time step, the electric field at each grid point can be calculated using the charge density at that grid point. The spatial electric field can be obtained by solving the Poisson equation, with the specific calculation formula as follows:

[0115] ;

[0116] in, Denotes the divergence operator, This represents the electric field strength.

[0117] Step A4: Calculate the electric field of the particle.

[0118] At each time step, an interpolation operation is required to map the electric field at the grid point to the particle location (the reverse of the process of calculating the charge density at the grid point).

[0119] ;

[0120] Represents the i-th particle electric field, Represents grid points The electric field.

[0121] Step A5: Particle propulsion.

[0122] At each time step, the particle's position and velocity are updated based on the calculated electric field. The Lorentz force acting on the particle and its motion are calculated using Newton's equations of motion:

[0123] ;

[0124] ;

[0125] and These represent the particle positions at step n and step (n+1) respectively. and Let represent the particle velocities at step n and step (n+1) respectively, m represent the particle mass, and q represent the particle charge. This represents the electric field at the particle; A time step is used to represent one motion of a particle over a time step.

[0126] Particle mesh methods approximate the phase space distribution using a large number of particles, and approximate the evolution of the phase space over time by observing the motion of particles. Although widely adopted, particle mesh methods face challenges of high memory consumption and computational complexity as computational scales increase (e.g., coupling tens of millions of particles with a 3D mesh or extreme resolutions). Specifically, particle mesh simulations require storing information such as particle position and velocity. As computational scales increase, the number of particles increases exponentially (each particle needs to be updated during the particle propagation step, and the computational load is positively correlated with the number of particles), with memory usage for particle information reaching terabytes (TB). Furthermore, the bidirectional iteration of particle motion and particle-field coupling, along with the solution of the Poisson equation, leads to high computational complexity and low efficiency.

[0127] As the computational scale increases, the number of particles used in particle meshes rises exponentially, leading to significant memory consumption and an exponential increase in computational cost. To maintain the computational scale (keeping the total number of particles constant) while reducing memory usage and computational cost, Singular Value Decomposition (SVD) is used as a new data structure to approximate the distribution of a large number of particles. Furthermore, a dynamic low-rank method and a basis-addition method are used as the motion update methods for this new data structure. To address the high memory and computational cost caused by the high particle number in particle mesh methods, this embodiment uses SVD to compress all particles and employs a dynamic low-rank method and a basis-addition method as new motion equations that evolve over time, thus accelerating the operation of the particle mesh method.

[0128] The low-rank representation of particle information (position and velocity) is as follows:

[0129] All particles in the particle grid are divided into one or more particle clusters based on the global or subdomain, and each particle cluster has its corresponding spatial matrix. and the corresponding velocity matrix The number of particles in each particle cluster is , The representation domain, where the number of particles in each particle cluster can be dynamically adjusted according to specific circumstances, describes the state of all particles in the corresponding region in the form of particle clusters. The spatial matrix X and the velocity matrix V are respectively decomposed into the product of three matrices using singular value decomposition:

[0130] ;

[0131] ;

[0132] Represents the space matrix corresponding to Orthogonal matrix;

[0133] Represents the space matrix corresponding to diagonal matrix;

[0134] Represents the space matrix corresponding to Orthogonal matrix;

[0135] Represents the transpose of a matrix;

[0136] Represents the velocity matrix corresponding to Orthogonal matrix;

[0137] Represents the velocity matrix corresponding to diagonal matrix;

[0138] Represents the velocity matrix corresponding to Orthogonal matrix.

[0139] By truncating the first k singular values ​​of the spatial matrix X, we can obtain a low-rank form of the spatial matrix (i.e., the first low-rank approximation matrix). By truncating the first j singular values ​​of the velocity matrix V, we can obtain a low-rank form of the velocity matrix (i.e., the second low-rank approximation matrix). .

[0140] Due to the truncation of singular values, and It became and orthogonal matrix, It became a diagonal matrix; and It became and orthogonal matrix, It became The diagonal matrix is ​​given below. The specific formula and explanation are as follows:

[0141] ;

[0142] ;

[0143] Represents the space matrix in low-rank form. Orthogonal matrix;

[0144] Represents the space matrix in low-rank form. diagonal matrix;

[0145] Represents the space matrix in low-rank form. Orthogonal matrix;

[0146] The velocity matrix in low-rank form corresponds to Orthogonal matrix

[0147] The velocity matrix in low-rank form corresponds to diagonal matrix;

[0148] The velocity matrix in low-rank form corresponds to Orthogonal matrix.

[0149] In the above formula, k is much smaller than the minimum value of a and b, and j is much smaller than the minimum value of a and b.

[0150] Through the above steps, the particle information (position and velocity) of the particle grid is represented as the product of three matrices. This not only ensures that the total number of particles remains unchanged, but also greatly compresses the data.

[0151] In this embodiment, the process of updating the velocity matrix based on the dynamic low-rank method is as follows:

[0152] Since the velocity matrix is ​​transformed into a low-rank approximation, the traditional Newtonian equations of motion are no longer applicable. Here, a dynamic low-rank approximation method is used as the new equations of motion to update the velocity matrix, by updating it separately. , and The matrix is ​​used to update the particle state. The specific process is as follows:

[0153] For solving differential equations:

[0154] , ;

[0155] This represents the value of the matrix at time t=0. This represents the initial matrix.

[0156] Let be a matrix of rank j. Its calculation process includes the following three sub-processes:

[0157] (1) Calculate using the Koopman-step (update the left singular matrix L) Increment :

[0158] , ;

[0159] connect and And perform QR decomposition:

[0160] , .

[0161] Koopman-step represents a method for transforming the dynamics of a nonlinear system into an infinite-dimensional linear system, and QR represents orthogonal decomposition.

[0162] (2) Calculate using Lie-Trotter-step (update right singular matrix R) Increment :

[0163] , ;

[0164] connect and And perform QR decomposition:

[0165] , ;

[0166] The Lie-Trotter-step method represents a way to break down complex differential equations into multiple simpler parts.

[0167] (3) S-step (updating the singular value matrix):

[0168] ;

[0169] The velocity matrix at this point (n+1 steps) for: S-step represents the number of steps to update the singular value matrix.

[0170] In the above text, , , , , and This represents the matrix generated during intermediate calculations;

[0171] This represents the electric field intensity matrix experienced by the particle;

[0172] Indicates the amount of charge on a particle;

[0173] Indicates the mass of the particle;

[0174] Indicates a time step;

[0175] The velocity matrix corresponding to the low-rank form of the nth step Orthogonal matrix;

[0176] The velocity matrix corresponding to the low-rank form of the nth step diagonal matrix;

[0177] The velocity matrix corresponding to the low-rank form of the nth step Orthogonal matrix;

[0178] This represents the velocity matrix in low-rank form corresponding to the (n+1)th step. Orthogonal matrix;

[0179] This represents the velocity matrix in low-rank form corresponding to the (n+1)th step. diagonal matrix;

[0180] This represents the velocity matrix in low-rank form corresponding to the (n+1)th step. Orthogonal matrix.

[0181] The process of updating the spatial matrix using the basis-addition method in this embodiment is as follows: Substitute the low-rank spatial matrix of step n and the low-rank velocity matrix of step (n+1) into the spatial equation to obtain the low-rank spatial matrix of step (n+1). , ;

[0182] Where n represents the number of iterations and the time step. Multiplying by a diagonal matrix reduces operations. Based on the summation form of the tensor product, the corresponding orthogonal matrix and diagonal matrix can be concatenated to obtain the following formula:

[0183] ;

[0184] so , and The matrix update result at step n+1 is as follows:

[0185] , , .

[0186] The space matrix corresponding to the low-rank form of the nth step Orthogonal matrix;

[0187] The space matrix corresponding to the low-rank form of the nth step diagonal matrix;

[0188] The space matrix corresponding to the low-rank form of the nth step Orthogonal matrix;

[0189] The space matrix corresponding to the low-rank form of the (n+1)th step Orthogonal matrix;

[0190] The space matrix corresponding to the low-rank form of the (n+1)th step diagonal matrix;

[0191] The space matrix corresponding to the low-rank form of the (n+1)th step Orthogonal matrix.

[0192] Both the dynamic low-rank method and the basis-increasing method calculate the increments of the three matrices separately, thereby updating the entire low-rank approximation. This method eliminates the need to update the state of each individual particle as in traditional particle mesh methods, instead updating the overall mode, thus satisfying the theoretical requirements for algorithm acceleration. This method is significantly faster than traditional particle mesh methods and its computational complexity does not increase with the number of particles. This embodiment uses a low-rank approximation form of singular value decomposition to represent particle information and employs the basis-increasing method and the dynamic low-rank method as the equations of motion for updating, achieving ultra-high data compression and reduced computational complexity.

[0193] The low-rank approximation simulation method based on particle mesh provided in this embodiment includes the following steps:

[0194] Step B1: Initialize the mesh and particles.

[0195] The physical region to be computed is divided into multiple uniform grids, each grid representing a small region of space; this step can generate all particles according to the computational scale and initial distribution, including the position and velocity of each particle; this step can also set the total number of iterations.

[0196] Step 2: Low-rank representation of particle information. Sort and integrate the positions and velocities of all particles into one or more... Given a spatial matrix X and a velocity matrix V, we obtain a low-rank approximation of the spatial matrix X and the velocity matrix V by applying singular value decomposition and truncating singular values ​​to X and V, respectively. and the velocity matrix in low-rank approximation form .

[0197] ;

[0198] ;

[0199] Step B3: Calculate the grid point charge density.

[0200] This step will Converted to a particle matrix, the particle matrix calculates the charge density at grid points using an interpolation function. .

[0201] ;

[0202] Step B4: Calculate the electric field at the grid points.

[0203] This step involves solving the Poisson equation using the charge density at the grid points to calculate the electric field at each grid point. The specific calculation formula is as follows:

[0204] .

[0205] Step B5: Calculate the electrostatic field experienced by the particle.

[0206] This step uses an interpolation function to map the electric field at the grid points to the particle positions. The calculation formula is as follows:

[0207] .

[0208] Step B6: Update the velocity matrix based on the dynamic low-rank method .

[0209] The electric field experienced by the particle is calculated in step B5. The velocity matrix at the nth step is calculated using the Koopman-step, Lie-Trotter-step, and S-step methods in the dynamic low-rank method. From the three submatrices, we can obtain the next step. .

[0210] Step B7: Update the space matrix using the basis addition method. .

[0211] Specifically, in this embodiment, the calculation can be obtained according to step B6. The space matrix at step n is updated using the basis-addition method. From the three submatrices, we can obtain the next step. :

[0212] Step B8: Map the particles based on periodic boundary conditions.

[0213] because After the update, some particles have moved out of the simulated physical space. This step uses periodic boundary conditions to remap these particles back into the physical space.

[0214] In this embodiment, the total number of iterations can be set. Steps B3 to B8 constitute one iteration, and the total number of iterations can be achieved by repeatedly executing steps B3 to B8.

[0215] This embodiment aggregates discrete particles in the particle mesh method into a matrix, and uses singular value decomposition and truncation of singular values ​​to obtain the corresponding low-rank form, which is then updated using a dynamic low-rank method. Compared with the traditional particle mesh method, the low-rank approximation matrix of this application not only ensures that the total number of particles remains unchanged, but also greatly compresses the amount of data, solving the problems of high memory consumption and data storage in large-scale simulations. This embodiment uses the dynamic low-rank method and basis-addition method to update the particles as a whole aggregated into the matrix, instead of updating the state of each individual particle. Therefore, the method in this embodiment is faster and solves the problem of a sharp increase in computational load caused by a high number of particles.

[0216] Please see Figure 2 , Figure 2 A flowchart of a low-rank approximation simulation method based on particle mesh provided for embodiments of this application is as follows:

[0217] S201: Initialize the network and particles.

[0218] S202: Determine the space matrix and velocity matrix in the low-rank approximation form.

[0219] S203: Calculate the point charge density of the grid.

[0220] S204: Solve the Poisson equation.

[0221] S205: Calculate the electric field experienced by the particle.

[0222] S206: Update the velocity matrix based on the dynamic low-rank method.

[0223] S207: Update the space matrix using the basis addition method.

[0224] S208: Mapping particles based on periodic boundary conditions.

[0225] S209: Determine if the iteration is complete; if yes, end the process; if no, proceed to S203.

[0226] While traditional particle mesh algorithms offer significant advantages in handling particle dynamics and electromagnetic field coupling problems, their computational complexity rapidly increases with the number of particles and the simulation dimension, leading to a significant computational bottleneck in large-scale applications. This embodiment addresses this by employing a novel data structure and ultra-high data compression capabilities. A new data structure obtained using singular value decomposition (SVD) represents a large number of particles, achieving compression capabilities tens of times greater, with the compression increasing with particle size. This embodiment utilizes dynamic low-rank methods and basis-addition methods for rapid computation of particle propagation steps. Compared to traditional particle meshes, this embodiment updates particle states from the overall modal perspective, rather than updating individual particles. Achieve speedup of more than 3 times in particle-scale scenarios.

[0227] This application provides a scene state update system, applied to an electronic device with singular value decomposition functionality. The system may include:

[0228] The particle mesh generation module is used to acquire scene parameters of the scene to be simulated and generate a corresponding particle mesh based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh is set with an arbitrary number of charged particles;

[0229] The matrix generation module is used to divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and to determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster.

[0230] The first processing module is used to perform singular value decomposition on the initial space matrix and truncate the singular value decomposition result of the initial space matrix to obtain a first low-rank approximate matrix.

[0231] The second processing module is used to perform singular value decomposition on the initial velocity matrix and truncate the singular value decomposition result of the initial velocity matrix to obtain a second low-rank approximate matrix.

[0232] The particle information update module is used to update the particle information of the particle grid using the first low-rank approximation matrix and the second low-rank approximation matrix;

[0233] The scene state update module is used to update the scene state of the scene to be simulated at the current moment based on the current particle information.

[0234] In this embodiment, after generating the particle mesh, all charged particles in the mesh are divided into an arbitrary number of particle clusters, and the initial spatial matrix and initial velocity matrix corresponding to each particle cluster are determined. This embodiment performs singular value decomposition and truncation operations on the initial spatial matrix and initial velocity matrix to obtain a first low-rank approximation matrix and a second low-rank approximation matrix. The truncation operation on the singular value decomposition results reduces the data dimensionality while retaining key information. The computation and storage cost of the low-rank approximation matrix is ​​much lower than that of the original high-dimensional matrix. The first and second low-rank approximation matrices can be used to quickly update the particle information of the particle mesh, thereby determining the scene state at the current moment. Therefore, this embodiment can reduce the computational load of the particle information update process and improve the computational efficiency of electronic devices.

[0235] Furthermore, the process by which the particle information update module updates the particle information of the particle grid using the first low-rank approximation matrix and the second low-rank approximation matrix includes: converting the first low-rank approximation matrix into a particle matrix; calculating the grid point charge density based on the particle matrix; solving the Poisson equation using the grid point charge density to obtain the grid point electric field; mapping the grid point electric field from the grid point to the position of the charged particle to obtain the electric field experienced by the charged particle; updating the second low-rank approximation matrix using the electric field experienced by the charged particle; updating the first low-rank approximation matrix using the updated second low-rank approximation matrix; and updating the particle information of the particle grid according to the updated first low-rank approximation matrix and the updated second low-rank approximation matrix.

[0236] Furthermore, the process by which the particle information update module updates the particle information of the particle grid based on the updated first low-rank approximation matrix and the updated second low-rank approximation matrix includes: determining the current position information of the charged particle based on the updated first low-rank approximation matrix; determining the current velocity information of the charged particle based on the updated second low-rank approximation matrix; and updating the particle information of the particle grid based on the current position information and the current velocity information.

[0237] Furthermore, it also includes:

[0238] The particle mapping module is used to determine whether the current position information of the charged particle is within the particle motion range after determining the current position information of the charged particle based on the updated first low-rank approximation matrix; if not, the charged particle is remapped into the particle motion range using periodic boundary conditions.

[0239] Furthermore, the process of the particle mesh generation module generating the particle mesh corresponding to the scene to be simulated includes: determining the physical space corresponding to the scene to be simulated, dividing the physical space into multiple uniform meshes, and generating an arbitrary number of charged particles in each mesh in order to generate the particle mesh.

[0240] Furthermore, the process by which the matrix generation module divides all the charged particles in the particle grid into an arbitrary number of particle clusters includes: dividing all the charged particles in the particle grid into one particle cluster based on the global domain; or, dividing all the charged particles in the particle grid into multiple particle clusters based on subdomains.

[0241] Accordingly, the process by which the matrix generation module determines the initial spatial matrix and initial velocity matrix corresponding to each particle cluster includes: determining the number of particles in the particle cluster; and generating the initial spatial matrix and initial velocity matrix corresponding to each particle cluster based on the number of particles.

[0242] Furthermore, the process of truncating the singular value decomposition result of the initial space matrix by the first processing module includes: determining a first rank value based on the number of rows and columns of the initial space matrix, and truncating the singular value decomposition result of the initial space matrix according to the first rank value; wherein, the first rank value is less than the number of rows of the initial space matrix, and the first rank value is less than the number of columns of the initial space matrix.

[0243] Accordingly, the process of truncating the singular value decomposition result of the initial velocity matrix by the second processing module includes: determining a second rank value based on the number of rows and columns of the initial velocity matrix, and truncating the singular value decomposition result of the initial velocity matrix according to the second rank value; wherein, the second rank value is less than the number of rows of the initial velocity matrix, and the second rank value is less than the number of columns of the initial velocity matrix.

[0244] This application also provides a scene state update system based on laser plasma instability distribution simulation, applied to electronic devices with singular value decomposition capabilities, including:

[0245] The particle mesh generation module is used to acquire scene parameters for simulating laser plasma instability distribution and generate corresponding particle meshes based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh contains an arbitrary number of charged particles.

[0246] The matrix generation module is used to divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and to determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster.

[0247] The first processing module is used to perform singular value decomposition on the initial space matrix and truncate the singular value decomposition result of the initial space matrix to obtain a first low-rank approximate matrix.

[0248] The second processing module is used to perform singular value decomposition on the initial velocity matrix and truncate the singular value decomposition result of the initial velocity matrix to obtain a second low-rank approximate matrix.

[0249] The particle information update module is used to update the particle information of the particle grid using the first low-rank approximation matrix and the second low-rank approximation matrix;

[0250] The scene state update module is used to update the scene state of the laser plasma instability distribution simulation at the current moment based on the current particle information.

[0251] Since the embodiments of the system part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the system part, and they will not be repeated here.

[0252] This application also provides a storage medium on which a computer program is stored, which, when executed, can perform the steps provided in the above embodiments. The storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0253] This application also provides an electronic device that may include a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it can implement the steps provided in the above embodiments. Of course, the electronic device may also include various network interfaces, power supplies, and other components.

[0254] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

[0255] It should also be noted that, in this specification, 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 a process, method, article, or apparatus. 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.

Claims

1. A scene state update method, characterized in that, Applications to electronic devices with singular value decomposition capabilities include: Obtain scene parameters of the scene to be simulated, and generate a corresponding particle mesh based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh is set with an arbitrary number of charged particles; Divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster. The initial spatial matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial spatial matrix is ​​truncated to obtain a first low-rank approximate matrix. The initial velocity matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial velocity matrix is ​​truncated to obtain a second low-rank approximate matrix. The particle information of the particle grid is updated using the first low-rank approximation matrix and the second low-rank approximation matrix; Update the scene state of the simulated scene at the current moment based on the current particle information; Updating the particle information of the particle mesh using the first low-rank approximation matrix and the second low-rank approximation matrix includes: The first low-rank approximation matrix is ​​converted into a particle matrix, and the grid point charge density is calculated based on the particle matrix. The Poisson equation is solved by the point charge density of the grid to obtain the electric field at the grid points; The electric field at the grid point is mapped from the grid point to the position of the charged particle to obtain the electric field experienced by the charged particle; The second low-rank approximation matrix is ​​updated using the electric field experienced by the charged particle; The first low-rank approximation matrix is ​​updated using the updated second low-rank approximation matrix; The particle information of the particle grid is updated based on the updated first low-rank approximation matrix and the updated second low-rank approximation matrix.

2. The scene state update method according to claim 1, characterized in that, The particle information of the particle mesh is updated based on the updated first low-rank approximation matrix and the updated second low-rank approximation matrix, including: The current position information of the charged particle is determined based on the updated first low-rank approximation matrix; The current velocity information of the charged particle is determined based on the updated second low-rank approximation matrix; The particle information of the particle mesh is updated based on the current position and velocity information.

3. The scene state update method according to claim 2, characterized in that, After determining the current position information of the charged particle based on the updated first low-rank approximation matrix, the method further includes: Determine whether the current position information of the charged particle is within the particle's motion range; If not, the charged particles are remapped into the particle motion range using periodic boundary conditions.

4. The scene state update method according to claim 1, characterized in that, Divide all the charged particles in the particle grid into an arbitrary number of particle clusters, including: All charged particles in the particle grid are divided into a particle cluster based on the global domain. Alternatively, all the charged particles in the particle grid can be divided into multiple particle clusters according to subdomains; Accordingly, the initial spatial matrix and initial velocity matrix corresponding to each particle cluster are determined, including: Determine the number of particles in the particle cluster; The initial space matrix and the initial velocity matrix are generated for each particle cluster based on the number of particles.

5. The scene state update method according to claim 1, characterized in that, The singular value decomposition result of the initial space matrix is ​​truncated, including: A first rank value is determined based on the number of rows and columns of the initial space matrix, and the singular value decomposition result of the initial space matrix is ​​truncated according to the first rank value; wherein, the first rank value is less than the number of rows of the initial space matrix and the first rank value is less than the number of columns of the initial space matrix; Accordingly, the singular value decomposition results of the initial velocity matrix are truncated, including: The second rank value is determined based on the number of rows and columns of the initial velocity matrix, and the singular value decomposition result of the initial velocity matrix is ​​truncated according to the second rank value; wherein the second rank value is less than the number of rows of the initial velocity matrix and the second rank value is less than the number of columns of the initial velocity matrix.

6. A scene state update method based on laser plasma instability distribution simulation, characterized in that, Applications to electronic devices with singular value decomposition capabilities include: Obtain scene parameters for simulating laser plasma instability distribution, and generate corresponding particle meshes based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh contains an arbitrary number of charged particles; Divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster. The initial spatial matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial spatial matrix is ​​truncated to obtain a first low-rank approximate matrix. The initial velocity matrix is ​​subjected to singular value decomposition, and the singular value decomposition result of the initial velocity matrix is ​​truncated to obtain a second low-rank approximate matrix. The particle information of the particle grid is updated using the first low-rank approximation matrix and the second low-rank approximation matrix; The simulation of the laser plasma instability distribution is updated based on the current particle information to reflect the current state of the scenario at the current moment. Updating the particle information of the particle mesh using the first low-rank approximation matrix and the second low-rank approximation matrix includes: The first low-rank approximation matrix is ​​converted into a particle matrix, and the grid point charge density is calculated based on the particle matrix. The Poisson equation is solved by the point charge density of the grid to obtain the electric field at the grid points; The electric field at the grid point is mapped from the grid point to the position of the charged particle to obtain the electric field experienced by the charged particle; The second low-rank approximation matrix is ​​updated using the electric field experienced by the charged particle; The first low-rank approximation matrix is ​​updated using the updated second low-rank approximation matrix; The particle information of the particle grid is updated based on the updated first low-rank approximation matrix and the updated second low-rank approximation matrix.

7. A scene state update system, characterized in that, Applications to electronic devices with singular value decomposition capabilities include: The particle mesh generation module is used to acquire scene parameters of the scene to be simulated and generate a corresponding particle mesh based on the scene parameters; wherein, the scene parameters include particle motion range, particle number, particle velocity and particle position, and each grid of the particle mesh is set with an arbitrary number of charged particles; The matrix generation module is used to divide all the charged particles in the particle grid into an arbitrary number of particle clusters, and to determine the initial spatial matrix and initial velocity matrix corresponding to each particle cluster. The first processing module is used to perform singular value decomposition on the initial space matrix and truncate the singular value decomposition result of the initial space matrix to obtain a first low-rank approximate matrix. The second processing module is used to perform singular value decomposition on the initial velocity matrix and truncate the singular value decomposition result of the initial velocity matrix to obtain a second low-rank approximate matrix. The particle information update module is used to update the particle information of the particle grid using the first low-rank approximation matrix and the second low-rank approximation matrix; The scene state update module is used to update the scene state of the scene to be simulated at the current moment based on the current particle information. The process by which the particle information update module updates the particle information of the particle grid using the first low-rank approximation matrix and the second low-rank approximation matrix includes: The first low-rank approximation matrix is ​​converted into a particle matrix, and the grid point charge density is calculated based on the particle matrix. The Poisson equation is solved using the grid point charge density to obtain the grid point electric field. The grid point electric field is mapped from the grid point to the position of the charged particle to obtain the electric field experienced by the charged particle. The second low-rank approximation matrix is ​​updated using the electric field experienced by the charged particle. The first low-rank approximation matrix is ​​updated using the updated second low-rank approximation matrix. The particle information of the particle grid is updated according to the updated first low-rank approximation matrix and the updated second low-rank approximation matrix.

8. An electronic device, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program, and the processor, when calling the computer program in the memory, implements the steps of the scene state update method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores computer-executable instructions, which, when loaded and executed by a processor, implement the steps of the scene state update method as described in any one of claims 1 to 6.

Images