Application-driven three-dimensional spatial data transmission method and system

Abstract

The present invention relates to the technical field of data transmission. Disclosed is an application-driven three-dimensional spatial data transmission method. The method comprises: determining system key performance indicators (KPIs) by means of qualitative and quantitative analysis; constructing an AI-driven adaptive three-dimensional data transmission mechanism, and dynamically adjusting a transmission strategy on the basis of a real-time network state, a device capability, an application scenario and the KPIs; developing an adaptive compression algorithm set oriented to three-dimensional data, so as to meet differentiated compression requirements of different application scenarios; performing loop execution of a test, and adjusting and optimizing the data transmission mechanism and the compression algorithm set on the basis of a test feedback result and the real-time network state; and deploying an optimized transmission method to a production environment, and collecting field data to optimize the system performance and verify the achievement of the KPIs. By constructing a qualitative and quantitative analysis framework based on machine learning, the present invention quantifies differentiated transmission requirements of different application scenarios and formulates transmission strategies meeting the scenario requirements.

Description

A three-dimensional spatial data transmission method and system driven by application Technical Field

[0001] The present invention relates to the field of data transmission technology, and in particular to an application-driven three-dimensional space data transmission method and system. Background Art

[0002] With the rapid development of emerging applications such as virtual reality (VR), augmented reality (AR), and 3D video, the real-time transmission of high-quality three-dimensional spatial data has become a critical capability. However, existing three-dimensional data transmission technologies have many shortcomings and cannot meet the performance requirements of new scenarios. Traditional network transmission mechanisms lack optimization for the characteristics of three-dimensional data and cannot efficiently transmit large and diverse three-dimensional data sets. Most three-dimensional compression algorithms focus on specific data formats (such as point clouds or meshes) and fail to provide universal adaptive compression support. In addition, static transmission strategies cannot be adjusted and optimized based on the dynamic changes in real-time network status and terminal device capabilities.

[0003] On the other hand, traditional network transmission mechanisms lack the ability to analyze application-specific requirements and model metrics, making it difficult to quantify and characterize the differentiated transmission requirements of different application scenarios. For example, for VR gaming, in addition to conventional metrics such as transmission latency and bandwidth usage, specialized indicators such as maintaining 3D data realism and visual continuity are also important. For industrial AR scenarios, key considerations include the accuracy of 3D interactions and model reconstruction time. Existing technologies lack scenario-specific KPI analysis and modeling frameworks, making it impossible to determine optimal transmission metrics for different scenarios and thus difficult to develop efficient transmission strategies. Furthermore, existing solutions lack the widespread application of artificial intelligence (AI), resulting in a lack of flexibility, adaptability, and continuous optimization capabilities.

[0004] Summary of the Invention

[0005] In view of the problems existing in the existing three-dimensional space data transmission technology, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to establish an intelligent three-dimensional data adaptive transmission mechanism based on artificial intelligence technology to meet the differentiated needs of different application scenarios and continuously optimize the transmission performance.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, an embodiment of the present invention provides an application-driven three-dimensional spatial data transmission method, which includes constructing an application scenario-oriented three-dimensional data transmission demand analysis system, and determining the system key performance indicators (KPIs) through qualitative and quantitative analysis; constructing an AI-driven adaptive three-dimensional data transmission mechanism, and dynamically adjusting the transmission strategy according to real-time network status, device capabilities, application scenarios and KPIs; developing an adaptive compression algorithm set for three-dimensional data to meet the differentiated compression requirements of different application scenarios; executing tests in a loop, and adjusting and optimizing the data transmission mechanism and compression algorithm set based on the test feedback results and real-time network conditions; deploying the optimized transmission method to the production environment, and collecting field data to optimize system performance and verify KPI achievement.

[0009] As a preferred solution of the application-driven three-dimensional spatial data transmission method described in the present invention, the following steps are included: determining the system key performance indicators KPIs through qualitative and quantitative analysis: based on machine learning and data mining technology, automatically extracting user demand indicators for three-dimensional data from historical usage data of various application scenarios; designing a qualitative and quantitative analysis framework based on artificial intelligence for the extracted demand indicators, and generating targeted evaluation strategies; establishing a scenario-based three-dimensional data transmission KPI model, mapping the demand indicators into measurable key performance indicators KPIs; integrating a data-driven simulation engine, simulating and replaying data for different networks, devices, and application scenarios based on the KPI model, and automatically determining the KPI threshold range for each scenario; the demand indicators include conventional indicators and characteristic indicators for three-dimensional spatial data, and the characteristic indicators for three-dimensional spatial data include three-dimensional data realism maintenance, hierarchical rendering quality, model reconstruction time, visual continuity maintenance, three-dimensional interaction accuracy, etc.

[0010] As a preferred solution of the application-driven three-dimensional spatial data transmission method described in the present invention, the following steps are included: if conventional performance indicators are measured, traditional network testing methods are used, and qualitative analysis is performed in combination with user experience feedback data; if the maintenance of three-dimensional data realism is measured, the degree of distortion of the three-dimensional model before and after compression / transmission is calculated based on the geometry and rendering similarity matrix, and the realism difference is analyzed using an image quality evaluation model based on deep learning, and the realism score is collected through subjective user testing, and the quantitative and qualitative evaluation results are integrated; if the hierarchical rendering quality is evaluated, the image quality defects are automatically identified and scored according to the rendering quality and the number of details at different resolutions, and an aesthetic ability model based on cloud rendering is deployed for the evaluation. The global image quality is evaluated, and real user satisfaction feedback is collected at the same time. If the model reconstruction time is measured, multiple test samples are tested in a real network environment and on real devices, the time required for complete reconstruction is recorded, and a timing model is used to predict the reconstruction time distribution in different scenarios to provide a confidence interval for the time evaluation. If visual continuity is evaluated, a visual fluency analysis model based on the attention mechanism is deployed to scan continuous frames, detect visual diffusion and frame skipping phenomena, and collect user visual rating feedback at the same time. A comprehensive evaluation is conducted based on the quantitative analysis results. If the accuracy of three-dimensional interaction is measured, an automated interaction evaluation platform based on computer vision and semantic analysis is designed to simulate real three-dimensional interaction scenarios, objectively evaluate the response time and the correctness of the results, and conduct qualitative analysis supplemented by user experience feedback.

[0011] As an optimal solution for the application-driven three-dimensional spatial data transmission method described in the present invention, the following steps are included: constructing an AI-driven adaptive three-dimensional data transmission mechanism: designing an adaptive transmission strategy model based on deep reinforcement learning, inputting real-time network status, device parameters and application scenarios, and autonomously generating the optimal strategy combination; the strategy combination includes data compression algorithm selection, resolution adjustment strategy, and transmission protocol stack parameters, supporting refined strategy search space and generating optimization combinations that surpass manual ones; deploying distributed optimization services that support heterogeneous computing power and environmental perception, perceiving network status changes in real time, and dynamically optimizing and scheduling transmission strategies in the cloud and at the edge; designing a user feedback collection module, optimizing the model with user feedback data on quality experience, and realizing continuous strategy optimization and model evolution.

[0012] As an optimal solution for the application-driven three-dimensional spatial data transmission method described in the present invention, the design of an adaptive transmission strategy generation model based on deep reinforcement learning includes the following steps: constructing a heterogeneous parallel decision network for parallel processing of multiple inputs including network status information, device parameter information and application scenario information; the heterogeneous parallel decision network includes a first subnetwork, a second subnetwork and a third subnetwork, the first subnetwork is a Transformer-based network state encoder for receiving the network status information and capturing its timing and topological characteristics; the second subnetwork is a graph neural network-based device topology encoder for receiving the device parameter information and mining the device topology information; the third subnetwork is a BERT-based application scenario semantic encoder for receiving the application scenario information and extracting scene semantic concepts; the outputs of the first subnetwork, the second subnetwork and the third subnetwork are fused into a unified environment state vector, and the environment state vector is input into a deep Q network; the deep Q network is used as a reinforcement learning agent to output the optimal strategy combination for three-dimensional data transmission based on the environment state vector.

[0013] As a preferred solution of the application-driven three-dimensional spatial data transmission method described in the present invention, the strategy combination includes data compression algorithm selection, resolution adjustment strategy, and transmission protocol stack parameters, including the following steps: the data compression algorithm selection strategy includes selecting a corresponding compression algorithm combination from an adaptive compression algorithm set according to the three-dimensional data type; adaptively adjusting the compression rate of the selected compression algorithm according to the key performance indicator KPI threshold range; the resolution adjustment strategy includes constructing a deep learning-based visual attention model to analyze the three-dimensional data to identify the key visual areas therein; for the key visual areas, improve their rendering resolution, and for the non-key visual areas, reduce their rendering resolution to achieve non-uniform resolution rendering; for continuous frame data, perform frame difference-based time sampling When there is no visual difference, the rendering resolution is reduced to reduce the rendering cost of visual redundancy; the transmission protocol stack parameter strategy includes determining the end-to-end configurable parameters including congestion control, error correction coding and packet header compression; based on the self-attention mechanism, the interaction between parameters is modeled. When the transmission overhead needs to be optimized, the congestion control parameters and packet header compression parameters are adjusted. When the transmission delay needs to be optimized, the error correction coding parameters are adjusted; the optimal combination of parameters is solved by the unconstrained optimization algorithm; the compression algorithm selection strategy, the resolution adjustment strategy and the transmission protocol stack parameter strategy constitute the strategy search space, whose size is the Cartesian product of each sub-strategy; a heterogeneous parallel decision network model is constructed to explore the strategy search space in real time; a strategy clipping algorithm based on heuristic rules is introduced to accelerate the convergence of strategy search to generate an optimization strategy combination that surpasses manual efforts.

[0014] As a preferred solution of the application-driven three-dimensional spatial data transmission method described in the present invention, the following comprises: developing an adaptive compression algorithm set for three-dimensional data, including using the idea of ​​generative adversarial networks (GANs) to model the compression process as an adversarial optimization process between a generator and a discriminator; designing a generator network structure, developing a customized encoder-decoder framework for different three-dimensional data representation forms, and introducing a self-attention mechanism to enhance the encoder-decoder framework's modeling ability for local and global features; designing a discriminator network structure, designing a Discriminator network based on 3D convolution or point cloud feature extraction, inputting original three-dimensional data and compressed reconstructed data, comparing the two and outputting a true / false score; designing a generative adversarial network training strategy, generating high-quality reconstruction and optimizing data discrimination capabilities by minimizing the discriminator score through the generator and integrating the compression rate target into the generator loss; using Set The Transformer or encoder-decoder attention structure models the dependencies between data points or feature channels through the self-attention mechanism; for different three-dimensional data representation forms, corresponding generator and discriminator models are designed respectively, and the self-attention mechanism is used to enhance the modeling capability; different generator and discriminator models are combined to form a family of three-dimensional data compression algorithms, and the corresponding family members are selected for compression according to the actual data format.

[0015] On the second aspect, an embodiment of the present invention provides an application-driven three-dimensional spatial data transmission system, which includes a demand analysis module for building an application scenario-oriented three-dimensional data transmission demand analysis system and determining the system's key performance indicators (KPIs) through qualitative and quantitative analysis; a strategy formulation module for building an AI-driven adaptive three-dimensional data transmission mechanism, dynamically adjusting the transmission strategy according to real-time network status, device capabilities, application scenarios and KPIs; a compression algorithm module for developing an adaptive compression algorithm set for three-dimensional data to meet the differentiated compression requirements of different application scenarios; a loop optimization module for looping tests and adjusting and optimizing the data transmission mechanism and compression algorithm set based on test feedback results and real-time network conditions; a deployment module for deploying the optimized transmission method to the production environment and collecting field data to optimize system performance and verify KPI achievement.

[0016] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the application-driven three-dimensional spatial data transmission method as described in the first aspect of the present invention are implemented.

[0017] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the application-driven three-dimensional spatial data transmission method as described in the first aspect of the present invention are implemented.

[0018] The beneficial effects of the present invention are as follows: the present invention constructs a qualitative and quantitative analysis framework based on machine learning to quantify the differentiated transmission requirements of different application scenarios and formulates transmission strategies that meet the scenario requirements; designs an adaptive transmission strategy generation model based on deep reinforcement learning to perceive the network, terminal and application scenario status in real time and dynamically generate the optimal strategy combination; develops an adaptive compression algorithm set for multiple three-dimensional data formats, supports automatic algorithm selection and online combination; adopts data-driven simulation and reinforcement learning technology to autonomously determine the KPI target threshold range in different scenarios; forms a closed-loop optimization mechanism, collects field data through stress testing and fault-tolerant testing, and realizes continuous optimization of transmission strategy models and compression algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] FIG1 is a flowchart showing the steps of an application-driven three-dimensional spatial data transmission method.

[0021] FIG2 is a diagram of computer equipment for an application-driven three-dimensional spatial data transmission method. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0025] Example 1

[0026] 1 and 2 , which are the first embodiment of the present invention, provide an application-driven three-dimensional spatial data transmission method, including:

[0027] S1: Build a three-dimensional data transmission demand analysis system oriented to application scenarios, and determine the system's key performance indicators (KPIs) through qualitative and quantitative analysis.

[0028] Specifically, the following steps are included:

[0029] S1.1: Based on machine learning and data mining technology, automatically extract user demand indicators for three-dimensional data from historical usage data of various application scenarios.

[0030] Preferably, the demand indicators include conventional indicators and characteristic indicators for three-dimensional spatial data. Conventional indicators include data transmission delay, throughput, data integrity, user experience, interactive response time and system scalability. Characteristic indicators for three-dimensional spatial data include maintenance of three-dimensional data realism, hierarchical rendering quality, model reconstruction time, visual continuity maintenance, and three-dimensional interaction accuracy.

[0031] S1.2: Design an AI-based qualitative and quantitative analysis framework for the extracted demand indicators and generate targeted evaluation strategies.

[0032] Specifically, if conventional performance indicators are measured, traditional network testing methods (ping test, iPerf and other tools to evaluate network indicators) are used, and qualitative analysis is performed in combination with user experience feedback data (APP ratings, feedback comments, etc.); if the realism maintenance of three-dimensional data is measured, the degree of distortion of the three-dimensional model before and after compression / transmission is calculated based on the geometry and rendering similarity matrix, and the image quality evaluation model based on deep learning is used to analyze the realism difference, and the realism score is collected through subjective user testing, and the quantitative and qualitative evaluation results are integrated; if the hierarchical rendering quality is evaluated, the image quality defects are automatically identified and scored according to the rendering quality and the number of details at different resolutions, and an aesthetic ability model based on cloud rendering is deployed to evaluate the global image quality. If the reconstruction time of the model is to be measured, multiple test samples are taken in a real network environment and on real devices, the time required for complete reconstruction is recorded, and a timing model is used to predict the distribution of reconstruction time in different scenarios to provide a confidence interval for the time evaluation; if visual continuity is to be evaluated, a visual fluency analysis model based on the attention mechanism is deployed to scan continuous frames, detect visual diffusion and frame skipping phenomena, and collect user visual rating feedback at the same time, and conduct a comprehensive evaluation based on the quantitative analysis results; if the accuracy of three-dimensional interaction is to be measured, an automated interaction evaluation platform based on computer vision and semantic analysis is designed to simulate real three-dimensional interaction scenarios, objectively evaluate the response time and the correctness of the results, and conduct qualitative analysis supplemented by user experience feedback.

[0033] S1.3: Establish a scenario-based three-dimensional data transmission KPI model and map demand indicators into measurable key performance indicators (KPIs).

[0034] S1.4: An integrated data-driven simulation engine simulates and replays different networks, devices, and application scenarios based on KPI models, automatically determining the KPI threshold range for each scenario.

[0035] Specifically, massive amounts of real network environment data, terminal equipment performance parameters, and application scenario data are collected and labeled to construct a three-dimensional data transmission environment dataset; a deep generative model is trained based on the three-dimensional data transmission environment dataset to simulate and restore the real three-dimensional data transmission environment, and support controllable adjustment of transmission parameters; the three-dimensional data transmission KPI model is integrated into the trained simulation engine, and the three-dimensional data stream is injected into the simulation environment to perform full-link KPI evaluation; the reinforcement learning model is trained using each KPI indicator as the environmental feedback signal and the adjustable transmission parameters as the behavior space of the reinforcement learning agent; the agent autonomously explores parameter combinations to determine the acceptable target threshold range of each KPI under different networks, devices, and application scenarios; historical real environment data is read, and offline KPI threshold pre-calculation and exploration are performed in the simulation environment; in an online environment, changes in the status of the network, equipment, and application scenarios are detected in real time, and the corresponding KPI thresholds are dynamically adjusted to achieve adaptive management.

[0036] S2: Build an AI-driven adaptive three-dimensional data transmission mechanism to dynamically adjust transmission strategies based on real-time network status, device capabilities, application scenarios, and KPIs.

[0037] Specifically, the following steps are included:

[0038] S2.1: Design an adaptive transmission strategy generation model based on deep reinforcement learning, input real-time network status, device parameters and application scenarios, and autonomously generate the optimal strategy combination.

[0039] Preferably, a heterogeneous parallel decision network is constructed for parallel processing of multiple inputs including network status information, device parameter information and application scenario information; the heterogeneous parallel decision network includes a first subnetwork, a second subnetwork and a third subnetwork, the first subnetwork is a Transformer-based network status encoder, which is used to receive network status information and capture its timing and topological characteristics; the second subnetwork is a graph neural network-based device topology encoder, which is used to receive device parameter information and mine device topology information; the third subnetwork is a BERT-based application scenario semantic encoder, which is used to receive application scenario information and extract scene semantic concepts; the outputs of the first subnetwork, the second subnetwork and the third subnetwork are fused into a unified environment state vector, and the environment state vector is input into the deep Q network; the deep Q network is used as a reinforcement learning agent to output the optimal strategy combination for three-dimensional data transmission based on the environment state vector.

[0040] S2.2: The strategy combination includes data compression algorithm selection, resolution adjustment strategy, and transmission protocol stack parameters, supporting a refined strategy search space and generating optimized combinations that surpass manual efforts.

[0041] Specifically, the data compression algorithm selection strategy includes, according to the three-dimensional data type, selecting the corresponding compression algorithm combination from the adaptive compression algorithm set; according to the key performance indicator KPI threshold range, adaptively adjusting the compression rate of the selected compression algorithm; the resolution adjustment strategy includes, building a visual attention model based on deep learning, analyzing the three-dimensional data to identify the key visual areas therein; for the key visual areas, improving the rendering resolution, and for the non-key visual areas, reducing the rendering resolution to achieve non-uniform resolution rendering; for continuous frame data, performing frame difference-based time sampling, and when there is no visual difference, reducing the rendering resolution to reduce the rendering cost of visual redundancy; the transmission protocol stack parameter strategy package Including, determining end-to-end configurable parameters including congestion control, error correction coding and packet header compression; based on the self-attention mechanism to model the interaction between parameters, when it is necessary to optimize the transmission overhead, adjust the congestion control parameters and packet header compression parameters, when it is necessary to optimize the transmission delay, adjust the error correction coding parameters; solve the optimal combination of parameters through an unconstrained optimization algorithm; the compression algorithm selection strategy, resolution adjustment strategy and transmission protocol stack parameter strategy constitute a strategy search space, whose size is the Cartesian product of each sub-strategy, which is beyond the range of manual enumeration; construct a heterogeneous parallel decision network model to explore the strategy search space in real time; introduce a policy clipping algorithm based on heuristic rules to accelerate the convergence of strategy search to generate an optimized strategy combination that surpasses manual efforts.

[0042] S2.3: Deploy distributed optimization services that support heterogeneous computing power and environmental awareness, perceive network status changes in real time, and dynamically optimize scheduling and transmission strategies in the cloud and edge.

[0043] S2.4: Design a user feedback collection module to loop user feedback data on quality of experience back to policy model training to achieve continuous policy optimization and model evolution.

[0044] S3: Develop a set of adaptive compression algorithms for 3D data to meet the differentiated compression requirements of different application scenarios.

[0045] Specifically, the following steps are included:

[0046] S3.1: Based on generative adversarial networks and self-attention mechanisms, design a family of compression algorithms that support multiple 3D data.

[0047] Using the idea of ​​generative adversarial networks (GANs), the compression process is modeled as an adversarial optimization process between a generator and a discriminator. The generator network structure is designed, and customized encoder-decoder frameworks are developed for different 3D data representation forms (such as point clouds, meshes, and volume data). The self-attention mechanism is introduced to enhance the encoder-decoder framework's modeling ability for local and global features. The discriminator network structure is designed, and the discriminator network is designed based on 3D convolution or point cloud feature extraction. The original 3D data and compressed reconstructed data are input, the two are compared, and a real / fake score is output. A generative adversarial network training strategy is designed to generate high-quality reconstructions and optimize data discrimination capabilities by minimizing the discriminator score through the generator and integrating the compression rate target into the generator loss. SetTransformer or encoder-decoder attention structure is used to model the dependency between data points or feature channels through the self-attention mechanism. Corresponding generator and discriminator models are designed for different 3D data representation forms, and the self-attention mechanism is used to enhance the modeling ability. Different generator and discriminator models are combined to form a family of 3D data compression algorithms, and the corresponding family members are selected for compression according to the actual data format.

[0048] The encoder-decoder framework consists of an encoder, which maps 3D data into a compact latent space representation, and a decoder, which reconstructs compressed 3D data from this latent space. The discriminator takes the original 3D data and the reconstructed compressed data as input, and its multi-scale and multi-view discriminator architecture facilitates the assessment of global and local quality differences.

[0049] S3.2: An evaluation module is built into the algorithm family to autonomously evaluate the conformity of compression results with KPIs and adaptively adjust the compression rate.

[0050] The evaluation formula for compliance S is as follows:

[0051] S=λ·Q+μ·C+v·E

[0052] Among them, λ, μ, and v represent weight coefficients, E represents the computational efficiency evaluation result, C represents the compression ratio evaluation result, and Q represents the reconstruction quality evaluation result. The specific formula is as follows:

[0053] Q = α·SSIM(D org ,D rec )+β·PSNR(D org ,D rec )+γ·(1-MHD(D org ,D rec ))

[0054] Among them, α, β, and γ represent weight coefficients, and D orgrepresents the original data, D rec Represents the reconstructed data, SSIM and PSNR measure the image quality, and MHD measures the similarity of point cloud shapes.

[0055] Among them, δ and υ represent weight coefficients, Size org Indicates the original data size, Size rec Indicates the compressed data size, D rec Represents the reconstructed data, ω i represents the weight of the i-th visual attention area, s i represents the subjective rating or objective quality score of the i-th visual attention region.

[0056] Among them, Time compress , Timedecompress represent the time required for compression and decompression respectively, Complexity (Ali) represents the complexity of the algorithm, ζ and η represent the weight coefficients.

[0057] S3.3: Design an automatic algorithm selection and online combination module to dynamically select and combine algorithm family members according to the strategy scheme to meet the compression requirements of different scenarios.

[0058] Specifically, a configuration profile is defined for each member algorithm in the algorithm family, including the supported three-dimensional data formats, achievable compression rate range, algorithm complexity level, etc.; an algorithm selection engine is developed, with the compression requirements of the strategy scheme as the engine input, and the algorithm that best meets the given requirements is selected from the algorithm family by applying heuristic rules or utilizing advanced machine learning models; if a single algorithm cannot meet the compression requirements, the algorithm combination strategy is applied to determine the cascade, parallel or hybrid combination method based on the given compression requirements; the algorithm selection engine is executed for each position in the combination, and the optimal algorithm combination scheme is determined based on compression quality, compression rate and complexity; an online combination execution engine is developed to receive input three-dimensional data in real time, and execute compression tasks on the pipeline according to the selected algorithm and its combination strategy; the compression result quality, compression rate and latency are collected through the feedback collection and optimization module. For compression results that do not meet the expected standards, the configuration is optimized and returned to the algorithm selection engine stage to reselect the algorithm. For new compression requirements, the configuration file is automatically expanded and the selection and combination are re-executed.

[0059] It should be noted that cascaded combinations are suitable for staged compression, selecting a different algorithm for each stage; parallel combinations are suitable for compressing the same dataset using different algorithms; and hybrid combinations combine cascaded and parallel combinations, using different combinations for different data subsets. The online combination execution engine supports dynamic loading and unloading of algorithm models.

[0060] S3.4: Deploy the algorithm family on a heterogeneous distributed platform to support distributed compression with cloud-edge collaboration.

[0061] S3.5: Improve algorithm performance through continuous learning, and support online access to new algorithms and automatic model tailoring.

[0062] S4: Execute the test in a loop and adjust and optimize the data transmission mechanism and compression algorithm set based on the test feedback results and real-time network conditions.

[0063] S5: Deploy the optimized transmission method to the production environment and collect field data to optimize system performance and verify KPI achievement.

[0064] Preferably, a network simulation tool is deployed to support simulation of network environments such as cellular networks, wired networks, and wireless networks, and to simulate different network conditions by adjusting bandwidth, latency, and packet loss rate; through end-to-end deployment of the optimized three-dimensional data transmission system, a full-link system test, including stress testing and fault tolerance testing, is performed; a stress test is performed to determine the upper limit of the system processing capacity. If the system processing capacity is insufficient, the failure phenomenon and context data are recorded, and the transmission strategy and compression algorithm are optimized in S2 or S3; a fault tolerance test is performed to verify the robustness of the system, simulating various abnormal scenarios. If the system cannot recover normally, the root cause of the failure is analyzed, and the fault tolerance and recovery mechanism is optimized; full-link field data is collected during the stress test and fault tolerance test, and the field data is analyzed using AI models to determine bottlenecks and locate potential problems, and the overall system is optimized based on the analysis results; based on the optimized system, the full-link test is re-performed to verify whether the previously set KPIs are met. If not, return to S1 to re-analyze the requirements and modeling; if it is met, the tested and optimized system is deployed to the production environment, and the changes in the network, equipment, and application scenarios are monitored in real time, and the transmission strategy and compression algorithm are dynamically adjusted.

[0065] Furthermore, this embodiment also provides a drive-oriented three-dimensional spatial data WEB transmission system, including a demand analysis module for building an application scenario-oriented three-dimensional data transmission demand analysis system, and determining the system's key performance indicators KPIs through qualitative and quantitative analysis; a strategy formulation module for building an AI-driven adaptive three-dimensional data transmission mechanism, and dynamically adjusting the transmission strategy according to real-time network status, device capabilities, application scenarios and KPIs; a compression algorithm module for developing an adaptive compression algorithm set for three-dimensional data to meet the differentiated compression requirements of different application scenarios; a loop optimization module for looping tests, and adjusting and optimizing the data transmission mechanism and compression algorithm set based on test feedback results and real-time network conditions; a deployment module for deploying the optimized transmission method to the production environment, and collecting field data to optimize system performance and verify KPI achievement.

[0066] This embodiment also provides a computer device suitable for the application-driven three-dimensional spatial data transmission method, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the application-driven three-dimensional spatial data transmission method proposed in the above embodiment.

[0067] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0068] This embodiment further provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the application-driven three-dimensional spatial data transmission method proposed in the above embodiment is implemented.

[0069] In summary, the present invention constructs a qualitative and quantitative analysis framework based on machine learning to quantify the differentiated transmission requirements of different application scenarios and formulate transmission strategies that meet the scenario requirements; designs an adaptive transmission strategy generation model based on deep reinforcement learning to perceive the network, terminal and application scenario status in real time and dynamically generate the optimal strategy combination; develops an adaptive compression algorithm set for multiple three-dimensional data formats, supporting automatic algorithm selection and online combination; adopts data-driven simulation and reinforcement learning technology to autonomously determine the KPI target threshold range in different scenarios; forms a closed-loop optimization mechanism, collects field data through stress testing and fault-tolerant testing, and realizes continuous optimization of transmission strategy models and compression algorithms.

[0070] Example 2

[0071] 1 and 2 , which are the second embodiment of the present invention, provide an application-driven three-dimensional spatial data transmission method. To verify the beneficial effects of the present invention, scientific demonstration is conducted through economic benefit calculations and simulation experiments.

[0072] Specifically, this simulation experiment was set up for a virtual reality (VR) gaming scenario. The 3D data used included a high-resolution game scene model consisting of 28,000 triangles and 50 million vertices. The experimental network environment simulated a 5G cellular network with a bandwidth of 200Mbps, a latency of 25ms, and a packet loss rate of 0.1%. The terminal device was an all-in-one VR headset equipped with a Snapdragon 865 processor, 8GB of RAM, and an Adreno 650 GPU.

[0073] Furthermore, based on the given network environment, terminal equipment and application scenarios, the simulation system simulates and infers the target range of key performance indicators (KPIs): data transmission delay <100ms, model reconstruction time <2s, and three-dimensional data realism maintenance score >4.8 (out of 5 points). After analyzing the current environmental status, the adaptive transmission strategy module generates the following optimization strategies: selecting a point cloud mesh compression algorithm, layered resolution rendering, enabling congestion control and packet header compression. Based on the strategy scheme, the compression algorithm module selects the GAN-driven point cloud mesh compression algorithm GAN-PC from the algorithm family and combines it with a layered resolution rendering algorithm based on visual attention. The compressed game scene model size is reduced from the original 6.3GB to 1.8GB, achieving a compression rate of 71.4%. Layered resolution rendering divides the screen into 5 levels, with the highest resolution being 2K and the lowest being 360P.

[0074] Furthermore, compressed three-dimensional data transmission was executed in a simulation environment while key indicators were monitored. The monitoring results showed: an average transmission delay of 84ms, an average model reconstruction time of 1.7s, and a realism maintenance score of 4.9 points, all of which met the expected KPI targets. At the same time, it was also found that some low-resolution areas had obvious image quality damage, and the user feedback scores were low. Based on the feedback, the system automatically adjusted the parameters of the layered resolution algorithm to improve the image quality details in low-resolution areas. The optimized transmission solution was re-executed in the simulation environment, and all indicators continued to meet the expected targets, and the user feedback scores were significantly improved.

[0075] Furthermore, the optimized transmission solution was deployed in the laboratory's real network environment (200Mbps bandwidth, 35ms latency, 0.2% packet loss rate) and real VR all-in-one equipment for stress testing. The stress test set a maximum of 30 concurrent people downloading game data at the same time to simulate peak scenarios. The test results showed: the average transmission delay was 115ms, the average reconstruction time was 2.1s, and the realism maintenance score was 4.7 points. Although some individual indicators were slightly lower than the simulation environment, the overall score was still within an acceptable range. The system supports horizontal expansion. When two new edge nodes were added to assist in transmission, the average delay dropped to 92ms, the reconstruction time was shortened to 1.8s, and the experience was further improved. The 30-hour long-term stability test showed no obvious performance degradation, confirming that the solution has good robustness. Based on test feedback, after continuous optimization of the system model, the three-dimensional data transmission solution passed verification and was put into a formal commercial environment.

[0076] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An application-driven three-dimensional spatial data transmission method, characterized by: include, Build a three-dimensional data transmission demand analysis system based on application scenarios, and determine the system's key performance indicators (KPIs) through qualitative and quantitative analysis; Build an AI-driven adaptive three-dimensional data transmission mechanism to dynamically adjust transmission strategies based on real-time network status, device capabilities, application scenarios, and KPIs; Develop an adaptive compression algorithm set for 3D data to meet the differentiated compression requirements of different application scenarios; Execute tests in a loop and adjust and optimize the data transmission mechanism and compression algorithm set based on test feedback and real-time network conditions; Deploy the optimized transmission method to the production environment and collect field data to optimize system performance and verify KPI achievement.

2. The application-driven three-dimensional data transmission method according to claim 1, wherein: Determining the system's key performance indicators (KPIs) through qualitative and quantitative analysis includes the following steps: Based on machine learning and data mining technology, it automatically extracts user demand indicators for 3D data from historical usage data of various application scenarios; Design an AI-based qualitative and quantitative analysis framework based on the extracted demand indicators and generate targeted evaluation strategies; Establish a scenario-based three-dimensional data transmission KPI model to map demand indicators into measurable key performance indicators (KPIs); An integrated data-driven simulation engine simulates and replays data for different networks, devices, and application scenarios based on KPI models, automatically determining the KPI threshold range for each scenario. The demand indicators include conventional indicators and characteristic indicators for three-dimensional spatial data. The characteristic indicators for three-dimensional spatial data include maintenance of three-dimensional data realism, hierarchical rendering quality, model reconstruction time, visual continuity maintenance, and three-dimensional interaction accuracy.

3. The application-driven three-dimensional data transmission method according to claim 2, wherein: Generating a targeted evaluation strategy includes the following steps: If measuring conventional performance indicators, traditional network testing methods are used, combined with qualitative analysis of user experience feedback data; To measure the realism of 3D data, the degree of distortion of the 3D model before and after compression / transmission is calculated based on the geometric and rendering similarity matrices, and the realism is analyzed using an image quality evaluation model based on deep learning. Differences were identified and realistic scores were collected through subjective user testing, combining quantitative and qualitative evaluation results; For graded rendering quality evaluation, image quality defects are automatically identified and scored based on the rendering quality and amount of detail at different resolutions. A cloud rendering-based aesthetic ability model is deployed to evaluate the overall image quality, while also collecting real-person user satisfaction feedback. If measuring model reconstruction time, test sampling multiple times in a real network environment and on real devices, record the time required for full reconstruction, and use a timing model to predict the distribution of reconstruction time in different scenarios to provide a confidence interval for the time estimate. To evaluate visual continuity, a visual fluency analysis model based on an attention mechanism is deployed to scan consecutive frames, detect visual dispersion and frame skipping, and collect user visual rating feedback, combining it with quantitative analysis results for comprehensive evaluation. To measure the accuracy of three-dimensional interaction, an automated interaction evaluation platform based on computer vision and semantic analysis should be designed to simulate real three-dimensional interaction scenarios, objectively evaluate response time and result correctness, and conduct qualitative analysis supplemented by user experience feedback.

4. The application-driven three-dimensional data transmission method according to claim 1, wherein: The construction of the AI-driven adaptive 3D data transmission mechanism includes the following steps: Design an adaptive transmission strategy generation model based on deep reinforcement learning, input real-time network status, device parameters and application scenarios, and autonomously generate the optimal strategy combination; The strategy combination includes data compression algorithm selection, resolution adjustment strategy, and transmission protocol stack parameters, supporting a refined strategy search space and generating an optimized combination that surpasses manual efforts; Deploy distributed optimization services that support heterogeneous computing power and environmental awareness, perceive network status changes in real time, and dynamically optimize scheduling and transmission strategies in the cloud and edge. Design a user feedback collection module to loop user feedback data on quality of experience back to policy model training to achieve continuous policy optimization and model evolution.

5. The application-driven three-dimensional data transmission method according to claim 4, characterized in that: The design of the adaptive transmission strategy generation model based on deep reinforcement learning includes the following steps: Build a heterogeneous parallel decision-making network to process multiple inputs in parallel, including network status information, device parameter information, and application scenario information; The heterogeneous parallel decision network includes a first sub-network, a second sub-network and a third sub-network. The first sub-network is a Transformer-based network state encoder, which is used to receive the network state information and capture its timing and topological characteristics; The second sub-network is a device topology encoder based on a graph neural network, which is used to receive the device parameter information and mine device topology information; The third sub-network is a BERT-based application scenario semantic encoder, which is used to receive the application scenario information and extract the scene semantic concepts; The outputs of the first sub-network, the second sub-network, and the third sub-network are combined into a unified environment state vector, and the environment state vector is input into the deep Q network; The deep Q network is used as a reinforcement learning agent to output the optimal strategy combination for three-dimensional data transmission based on the environment state vector.

6. The application-driven three-dimensional data transmission method according to claim 4, characterized in that: The strategy combination includes data compression algorithm selection, resolution adjustment strategy, and transmission protocol stack parameters, and includes the following steps: Data compression algorithm selection strategies include, Selecting a corresponding compression algorithm combination from an adaptive compression algorithm set according to the three-dimensional data type; Adaptively adjust the compression rate of the selected compression algorithm based on the key performance indicator (KPI) threshold range; The resolution adjustment strategy includes building a deep learning-based visual attention model to analyze 3D data to identify key visual areas. For key visual areas, Increase its rendering resolution, and reduce its rendering resolution for non-critical visual areas to achieve non-uniform resolution rendering; For continuous frame data, perform temporal sampling based on frame difference. When there is no visual difference, reduce the rendering resolution to reduce the rendering cost of visual redundancy. Transport protocol stack parameter strategies include: Determine end-to-end configurable parameters including congestion control, error correction coding, and packet header compression; Based on the self-attention mechanism, the interaction between model parameters is modeled. When transmission overhead needs to be optimized, congestion control parameters and packet header compression parameters are adjusted. When transmission delay needs to be optimized, error correction coding parameters are adjusted. Solve the optimal combination of parameters through unconstrained optimization algorithm; The compression algorithm selection strategy, resolution adjustment strategy and transmission protocol stack parameter strategy constitute the strategy search space, whose size is the Cartesian product of each sub-strategy; Build a heterogeneous parallel decision network model to explore the strategy search space in real time; Introducing a policy pruning algorithm based on heuristic rules to accelerate the convergence of policy search to generate strategies that surpass human Optimization strategy combination of workers.

7. The application-driven three-dimensional spatial data transmission method according to claim 6, characterized in that: The development of an adaptive compression algorithm set for three-dimensional data includes: Using the idea of ​​generative adversarial networks (GANs), the compression process is modeled as an adversarial optimization process between the generator and the discriminator. Design the generator network structure, Develop customized encoder-decoder frameworks for different 3D data representations, and introduce a self-attention mechanism to enhance the encoder-decoder framework's ability to model local and global features; Design the discriminator network structure, Design a Discriminator network based on 3D convolution or point cloud feature extraction, input the original 3D data and compressed reconstructed data, compare the two and output a true / false score; Design a generative adversarial network training strategy that produces high-quality reconstructions and optimizes data discrimination by minimizing the discriminator score through the generator and incorporating a compression rate objective into the generator loss; Use SetTransformer or encoder-decoder attention structure to model the dependencies between data points or feature channels through self-attention mechanism; For different 3D data representation forms, corresponding generator and discriminator models are designed respectively, and the self-attention mechanism is used to enhance modeling capabilities; Different generator and discriminator models are combined to form a family of three-dimensional data compression algorithms, and the corresponding family members are selected for compression according to the actual data format.

8. An application-driven three-dimensional spatial data transmission system, based on the application-driven three-dimensional spatial data transmission method according to any one of claims 1 to 7, characterized in that: Also includes, The demand analysis module is used to build a 3D data transmission demand analysis system oriented towards application scenarios and determine the system's key performance indicators (KPIs) through qualitative and quantitative analysis; A policy-making module, which is used to build an AI-driven adaptive three-dimensional data transmission mechanism and dynamically adjust the transmission strategy based on real-time network status, device capabilities, application scenarios, and KPIs; Compression algorithm module, used to develop adaptive compression algorithm sets for 3D data to meet differentiated compression requirements in different application scenarios; A loop optimization module is used to execute tests in a loop and adjust and optimize the data transmission mechanism and compression algorithm set based on the test feedback results and real-time network conditions; Deployment module, used to deploy the optimized transmission method to the production environment and collect field data for Optimize system performance and verify KPI achievement.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the application-driven three-dimensional spatial data transmission method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the application-driven three-dimensional spatial data transmission method according to any one of claims 1 to 7 are implemented.

Images