A rendering method and system based on digital twin display

Abstract

The present application provides a rendering method and system based on digital twin display, including: determining an area to be displayed, dividing the area to be displayed into grids to obtain multiple grids; allocating the multiple grids to available devices in a resource pool, so that the available devices use a trained reinforcement learning model to determine the optimal rendering strategy corresponding to each grid; using the optimal rendering strategy corresponding to each grid to render the multiple grids to obtain multiple rendered grids; integrating the multiple rendered grids to obtain the rendered area to be displayed. The present application divides the digital twin scene into multiple grids, and uses artificial intelligence technologies such as reinforcement learning to dynamically select the optimal rendering strategy for each grid, thereby realizing refined management and optimization of resources.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of digital twin technology, and in particular, to a rendering method and system based on digital twin display. Background Art

[0002] With the rapid development of digital twin technology, it has shown great potential in the fields of smart cities, smart manufacturing, smart healthcare, etc. Digital twins can not only simulate physical entities in the real world with high precision, but also achieve two-way interaction between the physical world and the virtual world through real-time data feedback, greatly improving the ability of decision support, remote monitoring and predictive analysis. However, with the increase in the complexity of digital twin models, especially the real-time rendering of large-scale scenes has become a technical challenge.

[0003] Traditional rendering methods often rely on fixed rendering strategies, such as uniform resolution, texture quality, and lighting effects. This "one-size-fits-all" approach is inadequate when faced with complex and ever-changing digital twin scenes, and can easily lead to waste or insufficient rendering resources. Summary of the invention

[0004] The embodiments of the present application provide a rendering method and system based on digital twin display to solve the technical problems existing in the above-mentioned background technology.

[0005] In the first aspect, an embodiment of the present application provides a rendering method based on digital twin display, including: determining an area to be displayed, dividing the area to be displayed into grids to obtain multiple grids; allocating the multiple grids to available devices in a resource pool, so that the available devices use a trained reinforcement learning model to determine the optimal rendering strategy corresponding to each grid, wherein the available devices include: a local computer, a workstation, an edge server, and a cloud-based GPU cluster; rendering the multiple grids using the optimal rendering strategy corresponding to each grid to obtain multiple rendered grids; integrating the multiple rendered grids to obtain a rendered area to be displayed.

[0006] Furthermore, the training process of the reinforcement learning model that has completed training is as follows: obtaining historical rendering data of the available device, wherein the rendering data includes: scene complexity score, user satisfaction score, rendering efficiency index and system resource utilization; preprocessing the historical rendering data to obtain a historical rendering data time series, wherein the preprocessing includes: data cleaning, standardization processing and feature engineering processing; using the historical rendering data time series, training the reinforcement learning model to obtain a reinforcement learning model that has completed training.

[0007] Furthermore, historical rendering data is obtained, including:

[0008] Obtaining scene complexity parameters, and using the scene complexity parameters to calculate the scene complexity score, wherein the scene complexity parameters include: the number of objects, the number of light sources, and the size of the user's attention area;

[0009] The calculation formula of the scene complexity score is:

[0010] ;

[0011] in, Score the complexity of the scene, to is the weight value, is the number of objects, is the number of light sources, is the size of the user's focus area, is the interaction distance between the user and the display area, is the attenuation factor, is a constant term;

[0012] Obtaining user satisfaction parameters, and using the user satisfaction parameters to calculate the user satisfaction score, wherein the user satisfaction parameters include: questionnaire satisfaction score, average usage time and number of repeated visits;

[0013] The calculation formula of the user satisfaction score is:

[0014] ;

[0015] in, rate the user's satisfaction, , and is the weight, Rate the satisfaction level of the questionnaire. For the duration of use, is the number of repeated visits;

[0016] Obtaining rendering efficiency index parameters, and using the rendering efficiency index parameters to calculate the rendering efficiency index, wherein the rendering efficiency index parameters include: frame rate, average rendering time and delay;

[0017] The calculation formula of the rendering efficiency index is:

[0018] ;

[0019] in, is the rendering efficiency index, to is the weight value;

[0020] Obtaining system resource utilization parameters, and using the system resource utilization parameters to calculate the system resource utilization, wherein the system resource utilization parameters include: GPU utilization, CPU utilization and memory occupancy;

[0021] The calculation formula of the system resource utilization is:

[0022] ;

[0023] in, is the system resource utilization, is the total number of the system resource utilization parameters, for System resource utilization parameters.

[0024] Furthermore, the reinforcement learning model is trained using the historical rendering data time series to obtain a completely trained reinforcement learning model, including: defining a vector composed of the historical rendering data time series as a state space, and defining adjustment of rendering parameters as an action space, wherein the rendering parameters include at least: resolution, texture quality, and shadow details; using the state space and the action space to train the reinforcement learning model to obtain a completely trained reinforcement learning model.

[0025] Furthermore, the reward function of the reinforcement learning model is:

[0026] ;

[0027] in, is the time step The reward value, , and is the weight value, , and The time The corresponding nonlinear conversion function of user satisfaction score, rendering efficiency index and system resource utilization, For time For the change in action at the previous time step, is the penalty coefficient, is a penalty term that decays over time, and Controls the initial strength and decay rate of the penalty, respectively.

[0028] Furthermore, allocating the multiple grids to available devices in a resource pool includes: calculating performance parameters between each grid and each available device, wherein the performance parameters include: time for the available device to render the grid, data transmission rate, energy consumption cost, and network stability; based on the performance parameters, calculating a comprehensive score between each grid and each available device; and based on the comprehensive score, allocating the multiple grids to corresponding available devices.

[0029] Furthermore, based on the performance parameters, a comprehensive score between each grid and each available device is calculated, including: calculating the inverse of the time for the available device to render the grid to obtain a first inverse, and calculating the inverse of the data transmission rate to obtain a second inverse; calculating the sum of the first inverse, the second inverse and the product of the network stability and the corresponding weight value, and using the difference between the sum of the products and the product of the energy consumption cost and the corresponding weight to obtain the comprehensive score.

[0030] In the second aspect, an embodiment of the present application provides a rendering system based on digital twin display, including: a division unit, used to determine an area to be displayed, and divide the area to be displayed into grids to obtain multiple grids; an allocation unit, used to allocate the multiple grids to available devices in a resource pool, so that the available devices can use the trained reinforcement learning model to determine the optimal rendering strategy corresponding to each grid, wherein the available devices include: a local computer, a workstation, an edge server, and a cloud-based GPU cluster; a rendering unit, used to render the multiple grids using the optimal rendering strategy corresponding to each grid to obtain multiple rendered grids; an integration unit, used to integrate the multiple rendered grids to obtain the rendered area to be displayed.

[0031] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the method described in the first aspect above.

[0032] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements the method described in the first aspect above.

[0033] In an embodiment of the present invention, by dividing the digital twin scene into multiple grids and using artificial intelligence technologies such as reinforcement learning, the optimal rendering strategy is dynamically selected for each grid, thereby achieving refined management and optimization of resources, thereby effectively solving the efficiency and quality balance problem faced by traditional rendering methods in large-scale and complex scenes, laying a solid foundation for the widespread application of digital twin technology, not only significantly improving the user experience, but also promoting efficient use of resources.

[0034] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 A flowchart of a rendering method based on digital twin display provided in an embodiment of the present application;

[0037] Figure 2 A schematic diagram of a rendering system based on digital twin display provided in an embodiment of the present application;

[0038] Figure 3 A schematic diagram of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Embodiment 1:

[0041] According to an embodiment of the present invention, an embodiment of a rendering method based on digital twin display is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0042] Figure 1is a flowchart of a rendering method based on digital twin display according to an embodiment of the present invention, such as Figure 1 As shown, the method comprises the following steps:

[0043] Step S102, determining a to-be-displayed area, and dividing the to-be-displayed area into grids to obtain a plurality of grids;

[0044] Step S104, allocating the multiple grids to available devices in a resource pool, so that the available devices use the trained reinforcement learning model to determine the optimal rendering strategy corresponding to each grid, wherein the available devices include: a local computer, a workstation, an edge server, and a cloud GPU cluster;

[0045] Step S106, rendering the plurality of grids using the optimal rendering strategy corresponding to each grid to obtain a plurality of rendered grids;

[0046] Step S108, integrating the rendered multiple grids to obtain a rendered area to be displayed.

[0047] The following will take a smart city digital twin display platform as an example to explain the above steps in detail.

[0048] First, the main areas of the city can be taken as the focus of display (i.e. the areas to be displayed), such as commercial areas, residential areas, industrial areas and park green spaces.

[0049] Next, the entire urban area is subdivided into thousands of grids, for example, each grid covers an area of ​​about 1 square kilometer. The size of the grid can be adjusted according to actual needs and computing resources to balance detail and performance.

[0050] Then, a resource pool was established, including high-performance GPU servers, edge computing nodes, and cloud computing resources. The most suitable rendering device is assigned to each grid using a pre-trained reinforcement learning model. The model considers the complexity of the grid (such as building density, pedestrian and vehicle traffic), user attention (such as the current online user's line of sight), and real-time environmental factors (such as weather changes) to determine the rendering strategy of each grid, such as texture resolution, shadow calculation, and dynamic effects.

[0051] Finally, after each grid is assigned to the corresponding device, it is rendered in real time according to its optimal rendering strategy. After rendering is completed, the images of all grids are seamlessly integrated to form a complete and coherent city digital twin view.

[0052] Through intelligent scheduling and dynamic adjustment of rendering strategies, the system can avoid unnecessary waste of resources while ensuring high-quality rendering of key areas, significantly improving overall performance and cost-effectiveness.

[0053] Users can enjoy a smooth and realistic digital twin experience, whether browsing the entire city layout at a macro level or examining the details of a single building at a micro level, they can get satisfactory visual effects and interactive experience.

[0054] The above flexibility allows it to easily adapt to changes in future urban development, whether it is new construction projects or more complex environmental simulation needs, which can be dealt with by adjusting the grid division and reinforcement learning model.

[0055] For urban planners and policymakers, the above digital twin demonstration provides unprecedented data analysis and visualization capabilities, which helps to make more informed, evidence-based decisions and promote sustainable development and intelligent management of cities.

[0056] In an embodiment of the present invention, the training process of the reinforcement learning model that has completed training is as follows:

[0057] Acquire historical rendering data of the available device, wherein the rendering data includes: scene complexity score, user satisfaction score, rendering efficiency index and system resource utilization;

[0058] Preprocessing the historical rendering data to obtain a time series of historical rendering data, wherein the preprocessing includes: data cleaning, standardization processing and feature engineering processing;

[0059] The reinforcement learning model is trained using the historical rendering data time series to obtain a reinforcement learning model that has completed training.

[0060] In the embodiment of the present invention, firstly, a large amount of historical rendering data is collected in the operation log of the digital twin display platform of the smart city, including:

[0061] Scene complexity score (C): Calculated based on indicators such as the number of objects in the scene, texture details, and lighting effects.

[0062] User satisfaction score (S): estimated through user feedback and interaction data, reflecting the user's subjective feeling about rendering quality.

[0063] Rendering efficiency index (E): covers parameters such as frame rate, rendering time and latency, reflecting the speed and smoothness of the rendering process.

[0064] System resource utilization (R): includes CPU, GPU, and memory usage, used to evaluate the impact of rendering tasks on system performance.

[0065] Specifically, obtain historical rendering data, including:

[0066] Obtaining scene complexity parameters, and using the scene complexity parameters to calculate the scene complexity score, wherein the scene complexity parameters include: the number of objects, the number of light sources, and the size of the user's attention area;

[0067] The calculation formula of the scene complexity score is:

[0068] ;

[0069] in, Score the complexity of the scene, to is the weight value, is the number of objects, is the number of light sources, is the size of the user's focus area, is the interaction distance between the user and the display area, is the attenuation factor, is a constant term;

[0070] Obtaining user satisfaction parameters, and using the user satisfaction parameters to calculate the user satisfaction score, wherein the user satisfaction parameters include: questionnaire satisfaction score, average usage time and number of repeated visits;

[0071] The calculation formula of the user satisfaction score is:

[0072] ;

[0073] in, rate the user's satisfaction, , and is the weight, Rate the satisfaction level of the questionnaire. For the duration of use, is the number of repeated visits;

[0074] Obtaining rendering efficiency index parameters, and using the rendering efficiency index parameters to calculate the rendering efficiency index, wherein the rendering efficiency index parameters include: frame rate, average rendering time and delay;

[0075] The calculation formula of the rendering efficiency index is:

[0076] ;

[0077] in, is the rendering efficiency index, to is the weight value;

[0078] Obtaining system resource utilization parameters, and using the system resource utilization parameters to calculate the system resource utilization, wherein the system resource utilization parameters include: GPU utilization, CPU utilization and memory occupancy;

[0079] The calculation formula of the system resource utilization is:

[0080] ;

[0081] in, is the system resource utilization, is the total number of the system resource utilization parameters, for System resource utilization parameters.

[0082] Next, the collected historical rendering data is preprocessed, including:

[0083] Data cleaning: Remove outliers and missing values ​​to ensure data integrity and accuracy.

[0084] Standardization processing: Convert numerical data to the same scale to eliminate the impact of dimension and facilitate model training.

[0085] Feature Engineering: Building more meaningful features, such as calculating the correlation between scene complexity score and user satisfaction, and the ratio of rendering efficiency to resource utilization.

[0086] Finally, use a deep reinforcement learning framework, such as DQN (Deep Q-Network) or PPO (Proximal Policy Optimization), as the base model.

[0087] The state space is defined as a vector consisting of scene complexity, user satisfaction, rendering efficiency, and resource utilization.

[0088] Define action space as adjusting rendering parameters such as resolution, texture quality, shadow detail, etc.

[0089] The reward function integrates parameters such as user satisfaction, rendering efficiency, and resource utilization to guide the model to learn to find the best balance between the three.

[0090] Preferably, the reward function is:

[0091] ;

[0092] in, is the time step The reward value, , and is the weight value, , and The time The corresponding nonlinear conversion function of user satisfaction score, rendering efficiency index and system resource utilization, For time For the change in action at the previous time step, is the penalty coefficient, is a penalty term that decays over time, and Controls the initial strength and decay rate of the penalty, respectively.

[0093] By using the historical rendering data time series and multiple iterative training, the model gradually learns to choose the optimal rendering strategy in different scenarios to maximize the value of the reward function.

[0094] The reinforcement learning model can dynamically adjust the rendering strategy according to the current scenario and user behavior, providing personalized high-quality rendering experience for different users at different times.

[0095] The reinforcement learning model can reasonably allocate system resources while ensuring rendering quality, avoiding resource waste caused by excessive rendering and improving the overall performance and cost-effectiveness of the platform.

[0096] Through continuous learning and optimization, the model can adapt to environmental changes, such as the evolution of user needs, hardware upgrades, and the introduction of new technologies, keeping the platform advanced and competitive.

[0097] The reinforcement learning model can quickly respond to complex and changing scenario requirements, provide data-driven decision support for platform managers, and improve operational efficiency and user experience.

[0098] In an embodiment of the present invention, allocating the plurality of grids to available devices in a resource pool includes:

[0099] Calculate the performance parameters between each grid and each available device, wherein the performance parameters include: the time for the available device to render the grid, the data transmission rate, the energy consumption cost and the network stability;

[0100] Based on the performance parameters, a comprehensive score between each grid and each available device is calculated;

[0101] Based on the comprehensive scores, the plurality of grids are allocated to corresponding available devices.

[0102] Specifically, based on the performance parameters, a comprehensive score between each grid and each available device is calculated, including:

[0103] Calculating the reciprocal of the time for the available device to perform grid rendering to obtain a first reciprocal, and calculating the reciprocal of the data transmission rate to obtain a second reciprocal;

[0104] The sum of the first inverse, the second inverse and the product of the network stability and the corresponding weight value is calculated, and the difference between the sum of the products and the product of the energy consumption cost and the corresponding weight is used to obtain the comprehensive score.

[0105] In the embodiment of the present invention, in the smart city digital twin display platform, in order to achieve efficient and resource-optimized rendering effects, we need to assign multiple grids in the digital twin scene to available devices in the resource pool for rendering. This process must not only take into account the rendering capabilities of the device, but also the data transmission efficiency, energy consumption cost and network stability to ensure that each grid can be rendered efficiently and with high quality on the most appropriate device.

[0106] Specifically, for each grid and every available device , calculate the following performance parameters:

[0107] Mesh rendering time available on the device , reflecting the available devices Rendering the Mesh The time required, including It can be expressed as the ratio between the complexity of the mesh and the rendering capabilities of the available device.

[0108] Data transfer rate , used to reflect the grid Data sources are transferred to available devices The speed of It can be expressed as the ratio between the physical distance between the data source and the available devices and the average transmission rate of the available devices.

[0109] Energy cost , available devices Rendering the Mesh energy consumption cost.

[0110] Network stability , reflecting the available devices Rendering the Mesh The stability of the network connection can be comprehensively evaluated based on indicators such as packet loss rate and delay change. High network stability means lower data transmission risks and a more stable task execution environment.

[0111] Finally, the reciprocal of the time for the available device to perform mesh rendering is calculated to obtain a first reciprocal, and the reciprocal of the data transmission rate is calculated to obtain a second reciprocal;

[0112] The sum of the first inverse, the second inverse and the product of the network stability and the corresponding weight value is calculated, and the difference between the sum of the products and the product of the energy consumption cost and the corresponding weight is used to obtain a comprehensive score.

[0113] Based on the comprehensive score of each mesh-device pair, the device with the highest comprehensive score is selected to render the corresponding mesh.

[0114] By comprehensively considering device rendering capabilities, data transmission efficiency, energy consumption costs and network stability, we ensure that each mesh can be rendered on the most appropriate device to maximize resource utilization efficiency.

[0115] The optimized grid-device allocation strategy can significantly shorten rendering time, increase data transmission speed, and reduce energy consumption costs, achieving dual optimization of performance and cost.

[0116] The introduction of network stability parameters ensures that even under poor network conditions, the device with the most stable network connection can be selected for rendering, improving the overall stability and reliability of the system.

[0117] Flexibility and adaptability: Based on the dynamic calculation of real-time performance parameters, this strategy can flexibly adapt to environmental changes, such as fluctuations in device performance and changes in network conditions, to ensure the best rendering effect in different scenarios.

[0118] Embodiment 2:

[0119] An embodiment of the present invention also provides a rendering system based on digital twin display, which is used to execute the rendering method based on digital twin display provided by the above content of the embodiment of the present invention. The following is a specific introduction to the rendering system based on digital twin display provided by an embodiment of the present invention.

[0120] like Figure 2 As shown, Figure 2 The figure is a schematic diagram of the above-mentioned rendering system based on digital twin display, which includes:

[0121] A division unit 10 is used to determine a to-be-displayed area and divide the to-be-displayed area into grids to obtain a plurality of grids;

[0122] The allocation unit 20 is used to allocate the plurality of grids to available devices in the resource pool, so that the available devices use the trained reinforcement learning model to determine the optimal rendering strategy corresponding to each grid, wherein the available devices include: a local computer, a workstation, an edge server, and a cloud GPU cluster;

[0123] A rendering unit 30, configured to render the plurality of grids using an optimal rendering strategy corresponding to each grid to obtain a plurality of rendered grids;

[0124] The integration unit 40 is used to integrate the rendered multiple grids to obtain a rendered area to be displayed.

[0125] In an embodiment of the present invention, by dividing the digital twin scene into multiple grids and using artificial intelligence technologies such as reinforcement learning, the optimal rendering strategy is dynamically selected for each grid, thereby achieving refined management and optimization of resources, thereby effectively solving the efficiency and quality balance problem faced by traditional rendering methods in large-scale and complex scenes, laying a solid foundation for the widespread application of digital twin technology, not only significantly improving the user experience, but also promoting efficient use of resources.

[0126] Embodiment three:

[0127] The embodiment of the present invention further provides a computing device for executing a program of the method described in the first embodiment, such as Figure 3 As shown, the computing device may include a storage component 41 and a processing component 42;

[0128] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 42 .

[0129] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the method of embodiment 1. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0130] The storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0131] The display component 43 may be an electroluminescent (EL) element, a liquid crystal display or a micro display having a similar structure, or a retinal direct display or a similar laser scanning display.

[0132] Of course, a computing device may also include other components, such as input / output interfaces, communication components, etc.

[0133] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0134] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0135] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0136] Embodiment 4:

[0137] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The method of the illustrated embodiment.

[0138] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0139] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0140] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A rendering method based on digital twin display, characterized in that: include: Determine an area to be displayed, and divide the area to be displayed into grids to obtain a plurality of grids; Allocating the multiple grids to available devices in a resource pool so that the available devices use the trained reinforcement learning model to determine the optimal rendering strategy corresponding to each grid, wherein the available devices include: a local computer, a workstation, an edge server, and a cloud GPU cluster, and the trained reinforcement learning model is a reinforcement learning model trained by historical rendering data of the available devices, and the rendering data includes: scene complexity score, user satisfaction score, rendering efficiency index, and system resource utilization; Rendering the plurality of grids using an optimal rendering strategy corresponding to each grid to obtain a plurality of rendered grids; Integrate the rendered multiple grids to obtain a rendered area to be displayed; Get historical rendering data, including: Obtaining scene complexity parameters, and using the scene complexity parameters to calculate the scene complexity score, wherein the scene complexity parameters include: the number of objects, the number of light sources, and the size of the user's attention area; The calculation formula of the scene complexity score is: ； in, Score the complexity of the scene, to is the weight value, is the number of objects, is the number of light sources, is the size of the user's focus area, is the interaction distance between the user and the display area, is the attenuation factor, is a constant term; Obtaining user satisfaction parameters, and using the user satisfaction parameters to calculate the user satisfaction score, wherein the user satisfaction parameters include: questionnaire satisfaction score, average usage time and number of repeated visits; The calculation formula of the user satisfaction score is: ； in, rate the user's satisfaction, , and is the weight, Rate the satisfaction level of the questionnaire. is the average usage time, is the number of repeated visits; Obtaining rendering efficiency index parameters, and using the rendering efficiency index parameters to calculate the rendering efficiency index, wherein the rendering efficiency index parameters include: frame rate, average rendering time and delay; The calculation formula of the rendering efficiency index is: ； in, is the rendering efficiency index, to is the weight value, is the frame rate; Obtaining system resource utilization parameters, and using the system resource utilization parameters to calculate the system resource utilization, wherein the system resource utilization parameters include: GPU utilization, CPU utilization and memory occupancy; The calculation formula of the system resource utilization is: ； in, is the system resource utilization, is the total number of the system resource utilization parameters, for System resource utilization parameters.

2. The method according to claim 1, characterized in that The training process of the reinforcement learning model that has completed the training is as follows: Acquire historical rendering data of the available device, wherein the rendering data includes: scene complexity score, user satisfaction score, rendering efficiency index and system resource utilization; Preprocessing the historical rendering data to obtain a time series of historical rendering data, wherein the preprocessing includes: data cleaning, standardization processing and feature engineering processing; The reinforcement learning model is trained using the historical rendering data time series to obtain a reinforcement learning model that has completed training.

3. The method according to claim 2, characterized in that The reinforcement learning model is trained using the historical rendering data time series to obtain a trained reinforcement learning model, including: Defining a vector composed of the time series of the historical rendering data as a state space, and defining adjusting rendering parameters as an action space, wherein the rendering parameters at least include: resolution, texture quality, and shadow details; The reinforcement learning model is trained using the state space and the action space to obtain a trained reinforcement learning model.

4. The method according to claim 3, characterized in that The reward function of the reinforcement learning model is: ； in, is the time step The reward value, , and is the weight value, , and The time The corresponding nonlinear conversion function of user satisfaction score, rendering efficiency index and system resource utilization, For time For the change in action at the previous time step, is the penalty coefficient, is a penalty term that decays over time, and Controls the initial strength and decay rate of the penalty, respectively.

5. The method according to claim 1, characterized in that Allocating the plurality of grids to available devices in a resource pool includes: Calculate the performance parameters between each grid and each available device, wherein the performance parameters include: the time for the available device to render the grid, the data transmission rate, the energy consumption cost and the network stability; Based on the performance parameters, a comprehensive score between each grid and each available device is calculated; Based on the comprehensive scores, the plurality of grids are allocated to corresponding available devices.

6. The method according to claim 5, characterized in that Based on the performance parameters, a comprehensive score is calculated for each grid and each available device, including: Calculating the reciprocal of the time for the available device to perform grid rendering to obtain a first reciprocal, and calculating the reciprocal of the data transmission rate to obtain a second reciprocal; The sum of the first inverse, the second inverse and the product of the network stability and the corresponding weight value is calculated, and the difference between the sum of the products and the product of the energy consumption cost and the corresponding weight is used to obtain the comprehensive score.

7. A rendering system based on digital twin display, used to execute the rendering method based on digital twin display according to any one of claims 1 to 6, characterized in that: include: A division unit, used for determining a to-be-displayed area, and dividing the to-be-displayed area into grids to obtain a plurality of grids; An allocating unit is used to allocate the plurality of grids to available devices in a resource pool, so that the available devices use the trained reinforcement learning model to determine the optimal rendering strategy corresponding to each grid, wherein the available devices include: a local computer, a workstation, an edge server, and a cloud GPU cluster; A rendering unit, used for rendering the plurality of grids by using an optimal rendering strategy corresponding to each grid to obtain a plurality of rendered grids; The integration unit is used to integrate the rendered multiple grids to obtain a rendered area to be displayed.

8. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the method according to any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the method according to any one of claims 1 to 6 is implemented.

Images