Method for dynamic management and compression of map spatial data fused with zpi and dcsa

Abstract

The application discloses a map space data dynamic management and compression method fusing ZPI and DCSA, and comprises the following steps: S1, map data is divided into n logical regions, and an inter-region adjacency graph is constructed; S2, a state vector is allocated to each region; S3, according to the state vector of each region, a region priority index ZPI is constructed, and regions are sorted according to the priority; S4, historical access data is collected, future access heat of the regions is predicted, and a region to be updated of a compression strategy is determined according to the access heat prediction result; an optimal compression strategy is obtained by using a dynamic compression scheduling algorithm DCSA and the access heat prediction result; and S6, map data loading is dynamically scheduled according to the priority index ZPI and the obtained optimal compression strategy. The application realizes dynamic update and adaptive sorting of region influence, and improves the foresight of scheduling of future hotspot regions.

Description

Technical Field

[0001] This invention relates to the field of spatial data management, and in particular to a method for dynamic management and compression of map spatial data that integrates ZPI and DCSA. Background Technology

[0002] With the rapid development of applications such as high-precision maps, digital twin cities, and connected vehicles, map spatial data is characterized by its massive scale, frequent updates, and complex access patterns. Traditional data management methods are gradually revealing a series of technical bottlenecks:

[0003] (1) Lack of dynamism in spatial index structure: Existing map systems typically use static spatial indexing methods such as R-Tree and quadtree. Although these methods have good performance in static data loading, they are difficult to adjust the indexing strategy in a timely manner when faced with frequent changes in the access popularity of a region (such as during peak commuting hours or when an emergency occurs), resulting in slow response in hotspot areas.

[0004] (2) The compression strategy is crude and lacks fine control: Most map service systems adopt a uniform compression level or a static compression mechanism based on layer resolution. This strategy fails to take into account factors such as user access behavior, regional functional attributes or data density. The compression strategy is out of touch with business needs, often resulting in missing information in hot areas or wasted resources in cold areas.

[0005] (3) Insufficient access prediction capability: Some systems support cache loading sorted by regional popularity, but they mostly rely on historical averages and do not introduce access trend modeling or dynamic learning mechanisms, so they cannot achieve advance scheduling of areas that are about to heat up.

[0006] (4) Lack of inter-regional association modeling mechanism: Most existing methods treat regions as independent units for management, without utilizing the inherent correlation between regions in terms of functional layout, traffic connection or user behavior, which limits the global optimization capability of index scheduling and caching strategies.

[0007] In summary, existing map spatial data management solutions generally suffer from problems of being "static, fragmented, and lagging," making it difficult to meet the comprehensive requirements of high dynamism, high real-time performance, and high precision control of map data in scenarios such as smart cities, real-time navigation, and emergency response. Summary of the Invention

[0008] To address the existing problems, this invention provides a method for dynamic management and compression of map spatial data that integrates ZPI and DCSA, the specific solution of which is as follows:

[0009] A method for dynamic management and compression of map spatial data integrating ZPI and DCSA includes the following steps:

[0010] S1, divide the map data into n logical regions V={ , ,…… , ..., Construct an adjacency graph between regions;

[0011] S2, for each region Assign state vectors, and for each region The initial state vector is: , which serve as input features for subsequently constructing the Region Priority Index (ZPI); where, For regional access frequency; Rate the feature tags; Historical scheduling priority; This represents the data density, which is the number of tiles divided by their size. For cache hit rate;

[0012] S3. Based on the state vectors of each region, construct the region priority index ZPI and sort the regions according to their priority.

[0013] S4: Collect historical access data, predict the future access popularity of the region, and determine the regions to be updated for the compression strategy based on the access popularity prediction results.

[0014] S5 uses the Dynamic Compression Scheduling Algorithm (DCSA) and access popularity prediction results to obtain the optimal compression strategy;

[0015] S6. Based on the priority index ZPI and the optimal compression strategy obtained in step S5, dynamically schedule the loading of map data.

[0016] Preferably, step S3 includes the following steps:

[0017] S31, using a graph convolutional network for multiple rounds of propagation: ;

[0018] in, Indicates the process Region after subconvolution propagation The state vector, Indicates the number of times the signal passes through the convolutional network; σ represents the attention weights between regions, calculated from geographical distance and event relevance; σ is the activation function. This is the layer weight matrix. Indicates the region The set of adjacent regions, Indicates the process The state vector of region j after propagation through the subconvolutional network; event association refers to the strength of association between regions due to real-time events, which is calculated as numerical weights through event co-occurrence and semantic similarity.

[0019] S32, the final output is the region priority index, ZPI: Used for dynamic sorting of loading and cache priority;

[0020] in, Indicates the region Priority rating, For normalized exponential functions, This is the transpose of the weight vector. This represents the region after propagation through an L-fold convolutional network. The state vector.

[0021] Preferably, step S4 specifically includes the following steps:

[0022] S41, Collect time-series access data , ,...., Input LSTM to predict the next stage of popularity: ;

[0023] in, Represents the region predicted by the LSTM model. Access popularity during the t+1 phase; arrive Indicates the region Historical access data from stage 1 to stage t.

[0024] S42, combined with current popularity, yields: Set the heat update threshold θ, when At that time, the area is marked as an area to be updated.

[0025] in, These are the weighting coefficients. This indicates the access popularity of region i after merging. It's about predicting the next phase of popularity. This reflects the current level of popularity.

[0026] Preferably, step S5 includes the following steps:

[0027] S51, Multidimensional Compression Strategy Space Construction:

[0028] The compression strategy for each region is defined as follows: ,in: For region i, the compression strategy is... Spatial resolution control level; Preserve hierarchy for semantic objects; Preserve frequency for time-dimensional data;

[0029] S52, Compression Cost Function Optimization:

[0030] Construct the compression cost function: ;in, This is the query delay time; This refers to the loss of accuracy caused by compression. α represents the required memory space; α, β, and γ are business sensitivity factors.

[0031] S53, searching for the optimal compression strategy using either a greedy or reinforcement learning strategy: ;

[0032] in, Let represent the optimal compression strategy for region i, argmin represent the variable value that minimizes the following expression, and Ω represent the set of compression strategies.

[0033] The optimal compression strategy Used to compress and reload the map data of region i, and to complete cache updates, rendering, switching and resource release.

[0034] The present invention also discloses a computer-readable storage medium storing a computer program, which, when executed, performs the method described in any of the above-mentioned embodiments.

[0035] The present invention also discloses a computer system including a processor, a storage medium storing a computer program, and the processor reading from the storage medium and running the computer program to perform the method described in any of the preceding claims.

[0036] The beneficial effects of this invention are as follows:

[0037] This invention introduces a graph neural network (GCN) to establish an inter-regional adjacency propagation mechanism, enabling dynamic updating and adaptive ranking (ZPI) of regional influence; it utilizes an LSTM sequence model to predict access popularity, improving the foresight of scheduling future hotspot areas; it constructs a three-dimensional compression strategy space of spatial precision, semantic redundancy, and temporal frequency, driving the search for the optimal compression strategy through a compression cost function; and it supports a refined resource scheduling strategy of on-demand compression + hot zone retention + cold zone unloading. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1This is a flowchart of the method of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] This invention relates to a method for efficiently managing and dynamically updating large-scale map spatial data by fusing graph neural indexing mechanisms and multidimensional compression control strategies.

[0042] The method of this invention is applicable to computing scenarios that require frequent access to map data and high real-time performance, such as smart cities, emergency dispatch, vehicle-to-everything (V2X) navigation, and 3D simulation.

[0043] Specifically, such as Figure 1 This invention discloses a dynamic management and compression method for map spatial data that integrates ZPI (Zone Priority Index) and DCSA (Dynamic Compression Scheduling Algorithm). ZPI stands for Zone Priority Index, and DCSA is a dynamic compression scheduling algorithm. The literature "Optimization of Big Data Parallel Scheduling Based on Dynamic Clustering Scheduling Algorithm" proposes the DCSA algorithm. This DCSA algorithm is a dynamic clustering scheduling algorithm based on data correlation, used for parallel optimization of big data tasks. The algorithm achieves optimal multi-channel scheduling through steps such as generating dynamic data queues, introducing weights for data balancing, and optimizing data placement based on data correlation. Experimental verification shows that it can achieve optimal scheduling of dynamic data, providing a solution to scheduling problems in big data processing.

[0044] The method includes the following steps:

[0045] S1 divides the map data into n logical regions. An adjacency graph is constructed between regions. The adjacency graph G=(V,E) is an undirected graph or a weighted directed graph, where the vertex set V corresponds to all logical regions, and the edge set E represents the adjacency relationship between regions. The adjacency relationship includes not only geographical boundaries but also logical adjacencies formed through transportation networks and functional connections. The edge weights can be initialized as the reciprocal of the geographical distance or assigned according to the strength of the association.

[0046] S2, for each region Assign state vectors, and for each region The initial state vector is: The state vector is used for graph convolutional network computation in subsequent step S3. For regional access frequency; Rate the functional tags (such as government affairs, business districts, hospitals, etc.); Historical scheduling priority; This represents the data density, which is the number of tiles divided by their size. This refers to cache hit rate. Among them, the feature tag score... It is a predefined numerical value used to characterize the static importance of a region. For example, a region can be labeled as a transportation hub, a business center, a residential area, a green space, etc., and each type can be assigned a base score (e.g., 0.9 for transportation hubs, 0.7 for business centers, 0.4 for residential areas, and 0.1 for green spaces). This score can be configured during system initialization.

[0047] S3. Based on the state vectors of each region, a Region Priority Index (ZPI) is constructed. Based on the region state vectors (access frequency, function tag ratings, etc.), a priority score is calculated using a graph neural network. This score is used to dynamically prioritize the loading and caching of map data, thus sorting regions according to priority. Specifically, this includes the following steps:

[0048] S31, using a graph convolutional network for multiple rounds of propagation: ;

[0049] in, Indicates the number of times the signal passes through the convolutional network; σ represents the attention weights between regions, calculated from geographical distance and event association; σ is the activation function (such as ReLU). This is the layer weight matrix; Describes the set of adjacent regions of region i. Indicates the process The state vector of region j after propagation through the subconvolutional network. "Event association" refers to the strength of the association between regions due to real-time events (such as traffic congestion or emergencies). It can be calculated as a numerical weight through event co-occurrence, semantic similarity, etc. The strength of event association can be based on information obtained from the real-time data interface. For example, if two regions are affected by the same emergency at the same time, or if there is a high-intensity traffic flow connection, then their event association weight is higher.

[0050] S32, the final output is the region priority index, ZPI: It is used for dynamic sorting of loading and caching priority.

[0051] S4, collect historical access data, predict future access popularity of the region, and determine the regions to be updated for the compression strategy based on the access popularity prediction results; the "access popularity" refers to the access popularity of the region (the value ranges from 0 to 1, and the higher the value, the greater the probability that the region is accessed), and the prediction results are used as the basis for strategy selection in the Dynamic Compression Scheduling Algorithm DCSA in step S5; specifically, this involves collecting time series access data. Input LSTM to predict the next stage of popularity: Input the LSTM model to predict the visit popularity in the next stage. Then, the current visit popularity is combined to obtain the final visit popularity.

[0052] Specifically, the following steps are included:

[0053] S41, Collect time-series access data Input LSTM to predict the next stage of popularity: ;

[0054] S42, combined with current popularity, yields: Set the heat update threshold θ, when At that time, the area is marked as an area to be updated. Among them, These are the weighting coefficients. This indicates the access popularity of region i after merging. It's about predicting the next phase of popularity. This reflects the current level of popularity.

[0055] S5. Using the Dynamic Compression Scheduling Algorithm (DCSA) and access popularity prediction results, the optimal compression strategy is selected. The optimal compression strategy refers to the compression strategy with the lowest cost selected by the compression cost function in the compression strategy space constructed based on the access popularity prediction results in three dimensions: spatial resolution, semantic object preservation, and temporal dimension preservation.

[0056] Specifically, the following steps are included:

[0057] S51, Multidimensional Compression Strategy Space Construction:

[0058] The compression strategy for each region is defined as follows: ,in: Spatial resolution control level; Preserve hierarchy for semantic objects; Preserve frequency for time-dimensional data. Spatial resolution control levels can be 0 (lossless), 1 (50% compression), 2 (75% compression), etc.; semantic object preservation levels can be 0 (retain all), 1 (retain only important features), 2 (retain only outlines), etc.; time-dimensional data preservation frequency can be 0 (real-time update), 1 (every 5 minutes), 2 (every hour), 3 (every day), etc. The specific level set Ω can be predefined according to business needs.

[0059] S52, Compression Cost Function Optimization:

[0060] Construct the compression cost function: ;in, This is the query delay time; This refers to the loss of accuracy caused by compression. α represents the required memory space; α, β, and γ are business sensitivity factors.

[0061] S53, searching for the optimal compression strategy using either a greedy or reinforcement learning strategy: .

[0062] The optimal compression strategy Used to compress and reload the map data for this area, and to complete cache updates, rendering, switching, and resource release.

[0063] S6 dynamically schedules map data loading according to priority indexing and optimal compression strategy.

[0064] During dynamic scheduling, the system maintains a priority queue. Data in regions with high ZPI scores and access frequency exceeding the threshold θ will be preferentially loaded into the cache with low compression ratio or lossless operation; data in regions with low ZPI scores and low access frequency will be compressed at a high ratio according to the optimal compression strategy or replaced from the cache to low-speed storage, thereby achieving fine-grained on-demand allocation of resources.

[0065] Example:

[0066] A city's traffic control center deployed a high-precision map service system, requiring the system to:

[0067] It displays real-time congestion status on main roads, bridges, and highways; automatically detects temporary road changes caused by construction or accidents; and maintains rapid response performance on mobile devices with limited resources.

[0068] The system map data is high-resolution tile data, covering approximately 220 square kilometers of the city, divided into 10m x 10m tile units. The total data volume reaches 15GB, with an average of over one million access requests per day.

[0069] The application steps of this invention in this scenario are as follows:

[0070] S1, Map region division and adjacency graph construction

[0071] (1) The map is divided into n=1024 logical regions;

[0072] (2) Establish a regional adjacency graph G=(V,E), where the adjacency edges are constructed by the connection relationship between the shared boundary and the main traffic artery;

[0073] S2, Initial state vector of each node: The construction results are shown in Table 1:

[0074] Table 1: Adjacency Graph Construction Results

[0075]

[0076] S3, ZPI priority evaluation in graph neural networks:

[0077] The weights are propagated using a 2-layer GCN network, and the ZPI priority ranking of each region is output.

[0078] Main road area ZPI i ≈ 0.93, marked as a level 1 cache;

[0079] ZPI in ordinary residential areas i ≈ 0.12, marked as Level 3 slow loading or compressed state.

[0080] S4, Access Popularity Prediction and Integration

[0081] S41 records hourly visits over the past 7 days. Construct an LSTM model:

[0082] During peak hours (7:00-9:00, 17:00-19:00) from Monday to Friday, the popularity index (T) on main roads is approximately 0.98; on weekends, the popularity index drops to around 0.33.

[0083] S42, LSTM predicts the popularity trend for the next day, and updates the regional popularity after incorporating current access data:

[0084]

[0085] The system sets the threshold θ=0.4, when At that time, the area is marked as needing to be updated.

[0086] S5, Dynamic Compression Strategy Scheduling

[0087] S51, based on fusion heat Execute compression strategy :

[0088] Main Road Area A (High Temperature Zone):

[0089] (Uncompressed);

[0090] (Keep all annotations);

[0091] (Updated every 5 minutes).

[0092] Residential Area B (Low Temperature Zone):

[0093] (Compression accuracy 50%)

[0094] (Only retain the main building labels);

[0095] (Updated daily only).

[0096] S52, for areas that meet the update conditions Enumerate a set of compression strategies And calculate the compression cost function:

[0097]

[0098] Choose the strategy with the lowest cost Applied to this area.

[0099] S53, System Performance Comparison (Experimental Results): As shown in Table 2:

[0100] Table 2: Comparison of System Performance Using the Method of the Invention and the Original System Method

[0101]

[0102] In summary, the map spatial management method based on regional priority indexing and dynamic compression scheduling disclosed in this invention includes: constructing a spatial adjacency graph, calculating ZPI priority using a graph neural network, dynamically adjusting the compression strategy based on access popularity prediction, and performing map data cache loading.

[0103] The Region Priority Index (ZPI) calculation of this invention is based on region access frequency, functional attributes, historical scheduling, data density, and cache hit rate, and uses a graph convolutional network for propagation learning. Furthermore, the compression strategy of this invention includes three aspects: spatial precision compression, semantic object sparsity, and temporal dimension preservation. The compression strategy is optimized and selected through a compression cost function. Simultaneously, the access popularity prediction of this invention uses an LSTM neural network for modeling, fusing historical and predicted values ​​to guide the dynamic switching of the compression strategy.

[0104] The following problems are solved by the solution of the present invention: (1) how to establish an adaptive and dynamically changing regional access priority mechanism; (2) how to implement a differentiated map data compression strategy based on factors such as access popularity, regional function, and semantic density; (3) how to achieve data offloading and lightweight storage of edge areas while ensuring the rendering accuracy of hot spots.

[0105] This invention introduces a graph neural network (GCN) to establish an inter-regional adjacency propagation mechanism, enabling dynamic updates and adaptive ranking of regional influence; it utilizes an LSTM sequence model to predict access popularity, improving the foresight of scheduling future hotspot areas; it constructs a three-dimensional compression strategy space of spatial precision, semantic redundancy, and temporal frequency, driving the search for the optimal compression strategy through a compression cost function; and it supports a refined resource scheduling strategy of on-demand compression + hot zone retention + cold zone unloading.

[0106] The present invention also discloses a computer-readable storage medium and a computer system. The computer-readable storage medium stores a computer program, which, upon execution, performs the method described in any of the preceding claims. A computer system includes a processor and a storage medium, on which a computer program is stored. The processor reads from the storage medium and runs the computer program to perform the method described in any of the preceding claims.

[0107] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability between hardware and software, the various illustrative components, blocks, modules, circuits, and steps are described above in a generalized manner in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in different ways for each specific application, but such implementation decisions should not be construed as departing from the scope of the invention.

[0108] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of both. The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor such that the processor can read and write information to / from the storage medium. In an alternative, the storage medium may be integrated into the processor. The processor and storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and storage medium may reside as discrete components in the user terminal.

[0109] In one or more exemplary embodiments, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functionality may be stored or transmitted as one or more instructions or code on or through a computer-readable medium. A computer-readable medium includes both computer storage media and communication media, encompassing any medium that facilitates the transfer of a computer program from one location to another. A storage medium may be any available medium accessible to a computer. By way of example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage, disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and is accessible to a computer. Any connection is also legitimately referred to as a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of a medium.

[0110] The prior description of this disclosure is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not intended to be limited to the examples and designs described herein, but should be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0111] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dynamic management and compression of map spatial data fusing ZPI and DCSA, characterized in that, The method comprises the following steps: S1, divide the map data into n logical regions , construct an inter-region adjacency graph; S2, for each region allocation status vector, and each region initialize the state vector to: where, is the region access frequency; is the function tag score; is the historical scheduling priority level; is the data density, i.e., tile quantity / size; is the cache hit rate; S3, constructing a region priority index ZPI according to the state vector of each region, and sorting the regions according to the priority; specifically comprising the following steps: S31, propagating through the graph convolution network for multiple rounds: ; wherein, represents the state vector of region j after propagation through the convolution network, represents the state vector of region j after propagation through the convolution network, represents the number of times of propagation through the convolution network; is the inter-region attention weight, calculated by geographical distance and event correlation; is an activation function; is a layer weight matrix, represents the set of adjacent regions of region j, represents the set of adjacent regions of region j, represents the state vector of region j after propagation through the convolution network, event correlation refers to the correlation strength between regions due to real-time events, which is calculated as a numerical weight through event co-occurrence and semantic similarity.​ S32, finally output region priority index ZPI: for dynamically ordering loading and cache priority; wherein, the priority score of a region, is a normalized exponential function, is the transpose of a weight vector, is a state vector of a region after L convolutional network propagations.​ S4, collecting historical access data, predicting the future access heat of the region, and determining the to-be-updated region of the compression strategy according to the access heat prediction result; S5, using a dynamic compression scheduling algorithm DCSA and the access heat prediction result to obtain an optimal compression strategy; specifically comprising the following steps: S51, multi-dimensional compression strategy space construction: Each region compression policy is defined as: where: is the compression policy for region i, is the spatial resolution control level; is the semantic object preservation level; is the time dimension data preservation frequency; S52, compression cost function optimization: Constructing compression cost function: ; wherein, is the query delay time; is the precision loss caused by compression; is the required memory space; and α, β, γ are service sensitive factors. S53, search the optimal compression strategy by greedy or reinforcement learning strategy: ; wherein, indicates the region of the optimal compression strategy, argmin indicates the variable value that makes the expression behind it take the minimum value, and Ω indicates the set of compression strategies; The optimal compression strategy For compressing and reloading the map data of a region , completing cache update, rendering, switching and resource release S6, dynamically scheduling map data loading according to the priority index ZPI and the obtained optimal compression strategy.

2. The method of claim 1, wherein, Step S4 specifically comprises the following steps: S41, collect time series access data Input LSTM to predict next stage heat: ; wherein, represents the region predicted by the LSTM model the visit heat at the t+1 stage; to represents the region the historical visit data at the 1 to t stages; S42, fusing the current heat to obtain: ; set a heat update threshold θ, when the region is marked as an area to be updated. wherein, is a weight coefficient, denotes the access heat of the fused region i, is the predicted next stage heat, is the current stage heat.

3. A computer-readable storage medium, characterized in that: The medium has a computer program stored thereon, and the computer program executes the method according to any one of claims 1 to 2 after running.

4. A computer system, characterized by: The method comprises a processor and a storage medium, the storage medium has a computer program stored thereon, and the processor reads and runs the computer program from the storage medium to execute the method according to any one of claims 1 to 2.

Images