A GIS application development method and system based on low code

Abstract

The present invention relates to the technical field of GIS application development, and discloses a low-code-based GIS application development method and system, comprising extracting spatial entities, spatial relationships, and spatial behaviors from real scenes, and constructing a demand relationship topology, inputting the demand relationship topology into a network analysis unit, and simultaneously orchestrating workflows, generating multiple candidate processes in parallel from the orchestrated workflows, simulating and deducing multiple solutions, selecting the optimal solution by prioritizing, and if multiple solutions do not meet the requirements, decomposing the orchestrated workflows into fingerprint high-order components, freely combining the fingerprint high-order components, abstracting the high-order components into topological embedding vectors, making components with similar functions adjacent to the topological embedding vectors in vector space, establishing a virtual model to simulate the combination of fingerprint high-order components and injecting virtual random disturbances, outputting the operation results of the high-order components, and triggering the simplification of the high-order components if the operation results fail, thereby achieving efficient GIS application development.

Description

Technical Field

[0001] The present invention relates to the technical field of GIS application development and discloses a low-code-based GIS application development method and system. Background Art

[0002] In current fields such as geographic information systems and intelligent decision-making systems, there are many challenges in spatial information processing and process optimization for real-world scenarios. Existing technologies often have problems with incomplete and inaccurate information extraction when extracting spatial entities, spatial relationships, and spatial behaviors, resulting in defects in the subsequent demand relationship topology. At the same time, in the process of workflow orchestration and scenario deduction, it is difficult to efficiently generate candidate processes in parallel. The simulation and deduction solutions are limited and lack a scientific prioritization mechanism, resulting in an inability to quickly find the optimal solution. In addition, when the solution does not meet the requirements, the existing technology lacks flexible and effective means of component disassembly, combination, and optimization, and cannot adapt to complex and changing business needs, which seriously restricts the decision-making efficiency and accuracy of the system. Summary of the Invention

[0003] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0004] To solve the above technical problems, the main purpose of the present invention is to provide a low-code-based GIS application development method, including:

[0005] Extract spatial entities, spatial relationships, and spatial behaviors from real-world scenarios, construct a demand relationship topology, input the demand relationship topology into a network analysis unit, and orchestrate workflows.

[0006] The orchestrated workflow generates multiple candidate processes in parallel, simulates and deduces multiple solutions, and selects the optimal solution by prioritization. If multiple solutions do not meet the requirements, the orchestrated workflow is disassembled into fingerprint high-level components;

[0007] Freely combine the fingerprint high-order components, abstract the high-order components into topological embedding vectors, and make components with similar functions adjacent to the topological embedding vectors in the vector space;

[0008] A virtual model is established to simulate the combination of fingerprint high-order components and inject virtual random disturbances, and the operation results of the high-order components are output. If the operation results fail, the simplification of the high-order components is triggered, and the constraints are adjusted to repair the failure of the results.

[0009] As a preferred solution of the low-code-based GIS application development method of the present invention, the following is a solution:

[0010] Extract multiple spatial entities from real-world scenes through multimodal interaction, obtain the longitude and latitude coordinates of the multiple spatial entities, and access the spatial database to assign spatial fingerprints to the spatial entities. The spatial fingerprints include type, location, and attributes.

[0011] The spatial relationship includes a topological relationship, a metric relationship, and a directional relationship. The topological relationship between multiple spatial entities is determined by the longitude and latitude coordinates of the spatial entities and the attributes of the spatial fingerprint. The distances between the multiple spatial entities are calculated from the longitude and latitude coordinates of the multiple spatial entities to determine the metric relationship. The directional relationship between the multiple spatial entities is determined by the vector direction of the longitude and latitude coordinate sequence.

[0012] The user operates the spatial entity to output an operation instruction, and the spatial behavior is extracted by parsing the operation instruction and binding the spatial fingerprint;

[0013] A demand relationship topology is constructed by the spatial entities, spatial relationships and spatial behaviors.

[0014] As a preferred solution of the low-code-based GIS application development method of the present invention, the following is a solution:

[0015] The demand relationship topology is imported into graph calculation, with spatial entities and spatial behaviors as heterogeneous nodes, and spatial relationships and logical dependencies as weighted edges, to construct the demand relationship topology.

[0016] The demand relationship topology is converted into a network model, and the node and edge information of the topology is serialized into a format compatible with the network analysis unit, and input into the network analysis unit. A triple association matrix is ​​constructed to convert the demand relationship topology into a computable feature. The triple association matrix includes an adjacency matrix, a metric matrix, and a direction matrix.

[0017] The network analysis unit analyzes the relationship topology and dynamically inserts a mutex to eliminate spatial conflicts and resource conflicts, and groups the behavior nodes without dependencies into the same level. The nodes in the same level are partitioned by spatial position and output to generate a directed acyclic graph with level numbering.

[0018] Each spatial behavior of the directed acyclic graph is converted into an executable unit.

[0019] As a preferred solution of the low-code-based GIS application development method of the present invention, the following is a solution:

[0020] The generation of multiple candidate processes includes: based on the feature representation of the demand relationship topology, searching historical process templates through a case matching engine, and directly reusing templates that meet similarity requirements; if there are no matching templates, launching a generative adversarial mechanism to create new candidate workflows and verify their logical completeness;

[0021] The candidate workflows are simulated and verified in a digital sandbox environment. By injecting multimodal disturbances such as simulated equipment failures and data anomalies, it is detected whether the output of the solution deviates from expectations. When the solution is marked as failed, the dynamic degradation mechanism is triggered, and the corresponding high-level components are disassembled into basic operation units. The degradation processing rules are refined collaboratively through the federated knowledge base to update the knowledge base.

[0022] As a preferred solution of the low-code-based GIS application development method of the present invention, the following is a solution:

[0023] The orchestrated workflow is a diversion engine, which generates multiple candidate processes in parallel. The diversion engine automatically selects and activates three types of generation strategies;

[0024] The three types of generation strategies described above retrieve the historical process template library to reuse workflows that meet the topological similarity requirements; apply predefined business logic rule chain reasoning to generate standard processes that comply with industry specifications; and use the generative adversarial mechanism to dynamically innovate and generate new workflows whose spatial logic rationality is verified.

[0025] As a preferred solution of the low-code-based GIS application development method of the present invention, the following is a solution:

[0026] Conduct multi-dimensional simulations of multiple candidate solutions generated in parallel in a digital sandbox environment. The sandbox dynamically injects multi-level disturbances such as data anomalies, equipment failures, and service conflicts to simulate the complexity of a real environment.

[0027] During the deduction process, the adaptive evaluation system collects data on the robustness, efficiency, cost, and user suitability of each solution in real time, generates a dynamic score, and filters out invalid solutions that are significantly inferior to other solutions in all evaluation dimensions, retaining the solution set that achieves the best balance in multiple dimensions. The system sorts the optimal solution set based on preset priority rules and user-defined preferences, and outputs the optimal solution for priority execution.

[0028] When all the candidate solutions selected through deduction fail to meet the core requirements, the system triggers the workflow intelligent disassembly mechanism.

[0029] As a preferred solution of the low-code-based GIS application development method of the present invention, the following is a solution:

[0030] The feature extraction engine performs multi-dimensional feature fusion on the functional semantics, topological associations, and industry attributes of fingerprint high-level components, and trains a topological embedding model based on a contrastive learning framework.

[0031] The topological embedding model maps components with similar functional semantics and topological roles to adjacent positions in the vector space.

[0032] As a preferred solution of the low-code-based GIS application development method of the present invention, the following is a solution:

[0033] The virtual model loads fingerprint high-order components and their topological embedding vectors through an adapter, and generates initial connection logic based on functional semantic similarity;

[0034] The virtual model also predefines the data layer, device layer, and business layer through a perturbation rule library, eliminating interference data in the database, device layer, and business layer through multi-level perturbations;

[0035] The adjacent fingerprint high-order components in the vector space are automatically concatenated through a topological semantic matching algorithm.

[0036] As a preferred solution of the low-code GIS application development system of the present invention,

[0037] When the running results deviate from the safety tolerance or the process is interrupted, the data is traced back to the failed high-level component, and the disassembler is called to deconstruct the failed high-level component into fine-grained basic units. At the same time, the disassembly depth is limited, the decomposition of key components is blocked, and a degradation fallback plan is generated;

[0038] Retesting whether the topology embedding vector library is complete by dynamically adjusting constraints, including loosening boundaries, parameter calibration, federation rule evolution, and incremental verification;

[0039] The relaxation bound is used for gradient extension failure constraint;

[0040] The parameter calibration adjusts input parameters based on feedback;

[0041] The federated rule evolution extracts similar failure strategies from a cross-domain knowledge base;

[0042] The incremental verification is used to retest whether the topology embedding vector library is constructed. If the construction is successful, the topology embedding vector library is updated to form a closed-loop optimization.

[0043] As a preferred solution of the low-code GIS application development system of the present invention,

[0044] Extract modules, arrange and disassemble modules, combine modules and virtual models;

[0045] The extraction module extracts spatial entities, spatial relations and spatial behaviors from real scenes and constructs demand relationship topology, and the network analysis unit choreographs the workflow;

[0046] The orchestration and disassembly module is used to generate multiple candidate processes in parallel from the orchestrated workflow, simulate and deduce multiple solutions, and disassemble the orchestration workflow into fingerprint high-level components;

[0047] The combination module freely combines the fingerprint high-order components, abstracts the high-order components into topological embedding vectors, and makes components with similar functions adjacent to the topological embedding vectors in the vector space;

[0048] The virtual model includes a component topology adapter, a perturbation rule base and dynamic constraints.

[0049] Beneficial effects of the present invention:

[0050] The present invention accurately extracts spatial entities, spatial relationships and spatial behaviors from real-world scenarios, constructs a demand relationship topology, inputs the demand relationship topology into a network analysis unit and orchestrates the workflow. It can generate multiple candidate processes in parallel and simulate and deduce them. Combined with a priority sorting mechanism, it can quickly screen out the optimal solution, significantly improving decision-making efficiency.

[0051] In this application, when multiple solutions do not meet the requirements, the orchestration workflow is decomposed into fingerprint high-level components, which are freely combined and abstracted into topological embedding vectors, so that similar functional components are adjacent in the vector space, which is convenient for management and calling, and enhances the flexibility and scalability of the system.

[0052] This application establishes a virtual model to simulate the combination of fingerprint high-level components and injects virtual random disturbances, which can detect potential problems in advance. When the operation results fail, the high-level components are triggered to simplify and adjust the constraint repair results, effectively improving the stability and reliability of the system, and providing an efficient and accurate solution for intelligent decision-making in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0054] Figure 1 This is a flowchart of a low-code-based GIS application development method of the present invention;

[0055] Figure 2 This is a composition diagram of a low-code-based GIS application development system of the present invention;

[0056] Figure 3 This is an overall simulation topology diagram of a low-code-based GIS application development method of the present invention;

[0057] Figure 4 This is a simulation diagram of a low-code-based GIS application development system of the present invention. DETAILED DESCRIPTION

[0058] 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.

[0059] 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.

[0060] 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.

[0061] Example 1

[0062] like Figure 1 As shown in the figure, a low-code-based GIS application development method includes:

[0063] Spatial entities, spatial relationships, and spatial behaviors are extracted from real-world scenarios, and a demand relationship topology is constructed. The demand relationship topology is input into a network analysis unit, and workflows are orchestrated simultaneously.

[0064] Through multimodal interaction, multiple spatial entities in the real scene are extracted, and the longitude and latitude coordinates of the multiple spatial entities are obtained. At the same time, the spatial fingerprint assigned to the spatial entity by accessing the spatial database includes type, location and attributes:

[0065] Geographic objects in real scenes are captured through multimodal interaction. Users draw or click directly on the map interface, or describe them through natural language. The system combines geocoding technology to convert semantics into longitude and latitude coordinates. At the same time, it accesses the spatial database to automatically match preset entities and identify elements such as building outlines and water body boundaries. All entities are given spatial fingerprints including type, location, and attribute sets, such as "Entity A: Type = Fire Station, Location = POINT (116.4, 39.9), Attribute = {Capacity: Medium}".

[0066] The spatial relationship includes a topological relationship, a metric relationship, and a directional relationship. The topological relationship between multiple spatial entities is determined by the longitude and latitude coordinates of the spatial entities and the attributes of the spatial fingerprint. The distances between the multiple spatial entities are calculated from the longitude and latitude coordinates of the multiple spatial entities to determine the metric relationship. The directional relationship between the multiple spatial entities is determined by the vector direction of the longitude and latitude coordinate sequence.

[0067] In this application, the directional relationship between multiple spatial entities includes topological relationship, metric relationship, and directional relationship.

[0068] Specifically, the topological relationship is determined based on the intersection logic of the entity boundary coordinates. For example, the spatial inclusion relationship is confirmed through the overlay analysis of administrative division polygons and park polygons.

[0069] Metric relationships calculate the true surface distance between entities, avoiding map projection distortion errors.

[0070] Directional relationships are determined by vector direction analysis of coordinate sequences, such as the derivation of river flow direction based on elevation attributes and coordinate orientation.

[0071] Spatial relationship extraction is based on the location of entities. The system uses spatial calculations to automatically deduce geometric relationships:

[0072] Specifically, topological relationships determine boundary intersection, such as checking whether a park is included in an administrative district;

[0073] Metric relationships calculate true surface distances between locations.

[0074] Directional relationships rely on vector analysis to determine azimuth.

[0075] For the semantic relationships described by users, the domain knowledge graph is called to map the influence into downstream diffusion paths in space.

[0076] The user operates the spatial entity to output an operation instruction, and the spatial behavior is extracted by parsing the operation instruction and binding the spatial fingerprint.

[0077] A demand relationship topology is constructed by the spatial entities, spatial relationships and spatial behaviors.

[0078] The specific demand relationship topology is constructed into a graph structure based on the extracted spatial entities, spatial relationships, and spatial behaviors, where entities are represented as nodes and relationships are represented as edges with attributes, for example, distance as edge weight and direction as edge label.

[0079] Furthermore, a data conversion module is used to serialize the node and edge information of the topology into a format compatible with the network analysis unit and input it into the unit in real time through a predefined API interface or message queue.

[0080] Furthermore, the network analysis unit performs operations such as shortest path calculation, community detection, or connectivity analysis through the connectivity of nodes and edges, thereby inferring spatial interaction patterns and discretizing continuous spatial problems into computable networks, achieving efficient spatial optimization and decision support, while ensuring performance through memory optimization and parallel computing when processing large-scale relationships.

[0081] The demand relationship topology is imported into graph calculation, with spatial entities and spatial behaviors as heterogeneous nodes, and spatial relationships and logical dependencies as weighted edges, to construct the demand relationship topology.

[0082] When the topology is input, the workflow orchestration system is activated to manage the end-to-end process. The orchestrator defines the task sequence, including topology input, network analysis execution, result output, and user feedback loop; for example, it dynamically schedules network analysis tasks through an event-driven mechanism and handles dependencies.

[0083] Furthermore, the workflow maintains the task state machine to ensure retry or rollback when the analysis fails, and integrates the input of the spatial database and user operation flow. At the same time, queue buffering and priority scheduling handle real-time interaction, ultimately achieving robustness and low latency in scene response to meet the needs of dynamic real-world applications.

[0084] In this application, a preferred specific implementation method of the demand relationship topology includes:

[0085] The demand relationship topology can be converted into a network model, and the node and edge information of the topology can be serialized into a format compatible with the network analysis unit and input into the network analysis unit to construct a triple association matrix, which includes an adjacency matrix, a metric matrix and a direction matrix.

[0086] The adjacency matrix is ​​used to describe the existential associations between entities. It encodes the basic connectivity between entities through 0 / 1 binary values, such as whether a road is connected to a building. If all rows / columns are 0, it is an isolated entity to exclude invalid behavior.

[0087] The metric matrix is ​​used to quantify the strength of spatial constraints, map the interaction costs between entities through the Euclidean distance calculation method, provide conflict judgment thresholds for the conflict resolution mechanism, and provide partitioning basis for the parallelism optimizer.

[0088] The direction matrix is ​​used to capture the directional constraints of spatial behavior and drives subsequent operations by encoding the directional dependence of the behavior with a unit vector.

[0089] The orchestrated workflow generates multiple candidate processes in parallel, simulates and deduces multiple solutions, and selects the optimal solution through priority sorting. If multiple solutions do not meet the requirements, the orchestrated workflow is disassembled into fingerprint high-level components.

[0090] The generation of multiple candidate processes includes: based on the feature representation of the demand relationship topology, retrieving historical process templates through the case matching engine, and directly reusing them when there are templates that meet the similarity requirements; if there are no matching templates, starting the generative adversarial mechanism to create a new candidate workflow and verify its logical completeness.

[0091] Specifically, parallel process generation is based on the topological graph feature vector, and historical templates are retrieved through the case matching engine. When there is no match, the generative adversarial network is activated: the generator produces candidate workflows, and the discriminator verifies the logical completeness.

[0092] In this application, a preferred method for establishing a countermeasure mechanism includes:

[0093] When network analysis fails, it automatically triggers a retry or rollback to the state before topology input. The real-time geographic data stream of the spatial database and the user operation event stream use priority message queues to buffer high-concurrency requests and dynamically schedule tasks according to the urgency of the instructions. In the demand relationship topology implementation, the topology converter serializes the graph structure topology into a network analysis unit compatible format and constructs a triple association matrix.

[0094] The adjacency matrix encodes the connectivity between entities as 0 / 1 binary values, such as road-building connections, and all-zero rows / columns identify isolated entities to exclude invalid behaviors.

[0095] The metric matrix quantifies the strength of spatial constraints, such as logistics costs, and issues an alarm when the distance is less than a safety threshold. The output value drives conflict resolution and parallel optimization, allowing multi-node concurrency when the distance is greater than the parallel threshold.

[0096] The direction matrix encodes the behavioral direction constraints through unit vectors, defines the timing execution order, and provides a benchmark for direction mutation detection.

[0097] Based on matrix feature representation, parallel process generation is activated by intelligent branching.

[0098] If there is no match in the adversarial process, the generative adversarial mechanism is activated to create candidate flows and verify the rationality of the solution based on the spatial logic rule base.

[0099] When the digital sandbox test marking scheme fails, high-level components are disassembled and simplified restrictions are imposed to generate degradation fallback plans. Finally, a multi-dimensional equilibrium solution set is screened to achieve a closed loop from topology construction to optimal execution.

[0100] The candidate workflows are simulated and verified in a digital sandbox environment. By injecting multimodal disturbances such as simulated equipment failures and data anomalies, it is detected whether the output of the solution deviates from expectations.

[0101] When a solution is marked as invalid, the dynamic degradation mechanism is triggered, the corresponding high-level components are disassembled into basic operation units, and the degradation processing rules are refined through the federated knowledge base to update the knowledge base.

[0102] Specifically, a digital sandbox verifies the injected multimodal perturbation parameters.

[0103] In the digital sandbox environment, the system dynamically generates multi-level perturbation parameters covering the data layer, device layer and business layer through the perturbation rule library to holographically simulate the complexity of real-world scenarios.

[0104] Data layer perturbations can use probabilistic models to generate random coordinate offsets, simulate GPS signal interference, and verify the spatial analysis's tolerance to positioning drift. They can also randomly mask key entity attributes at preset ratios to test the solution's adaptability to data loss.

[0105] Inject out-of-order and repeated data streams reported by sensor anomalies to verify the robustness of timing analysis and map hardware resource anomaly scenarios with device-layer disturbances.

[0106] Specific device-layer disturbance mapping hardware resource anomaly scenarios are used to simulate edge node memory overflow and CPU overload to trigger computing degradation strategies, and periodically block network links to test offline decision-making capabilities.

[0107] Business layer disturbances are implanted into logical conflict events to construct mutually exclusive operation instructions, such as requesting the demolition and protection of the same building at the same time.

[0108] The verification conflict resolution mechanism is used to forge unauthorized operations to check the effectiveness of security interception. The verification conflict resolution mechanism is used to, for example, trigger the control process by a low-privileged user.

[0109] Perturbation injection follows the principle of scenario weight adaptation. When the sandbox detects that the output deviates from the expectation by more than the tolerance threshold, the scheme is marked as invalid, and then atomic degradation and disassembly are triggered, deconstructing the failed high-order components into basic operation units while imposing disassembly constraints.

[0110] Collect degradation cases in real time, establish a new strategy disturbance rule base, and form a self-reinforcing mechanism for disturbance generation, failure detection, degradation processing, and knowledge evolution.

[0111] After verification in the rail transit dispatching scenario, after injecting disturbances such as positioning drift and communication interruption, component degradation was completed and a new signal loss-enabled offline timetable strategy was added to improve the resilience of the low-code GIS application development system.

[0112] Degradation cases can extract sensor failover interpolation algorithm rules across projects.

[0113] The orchestrated workflow is a diversion engine, which generates multiple candidate processes in parallel. The diversion engine automatically selects and activates three types of generation strategies;

[0114] Specifically, the multi-strategy parallel candidate generation process automatically selects and activates three types of generation strategies based on the structural characteristics of the demand topology graph, such as node connection density and relationship type distribution.

[0115] The three types of generation strategies described above retrieve the historical process template library to reuse workflows that meet the topological similarity requirements; apply predefined business logic rule chain reasoning to generate standard processes that comply with industry specifications; and use the generative adversarial mechanism to dynamically innovate and generate new workflows whose spatial logic rationality is verified.

[0116] Retrieve workflows for similar scenarios from the template library, control reuse conditions through topological similarity thresholds, apply predefined business logic rules, and generate standard processes that comply with industry specifications. For example, monitoring pollution requires first defining a buffer zone.

[0117] Through the generative adversarial mechanism, the generator creates a new workflow and the discriminator verifies its spatial logical rationality.

[0118] Multi-dimensional simulation and deduction are performed on multiple candidate solutions generated in parallel in a digital sandbox environment. The sandbox dynamically injects multi-level disturbances such as data anomalies, equipment failures and business conflicts to simulate the complexity of the real environment.

[0119] During the deduction process, the adaptive evaluation system collects the robustness, efficiency, cost and user adaptability of each solution in real time to form a dynamic score. By screening out invalid solutions that are inferior to other solutions in all evaluation dimensions, the system retains the set of solutions that achieve the best balance in multiple dimensions. The system sorts the optimal solution set according to preset priority rules and user-defined preferences, and outputs the optimal solution for priority execution.

[0120] When all the candidate solutions that have been optimized through deduction fail to meet the requirements, the system triggers the workflow intelligent disassembly mechanism. Specifically, the intelligent disassembly mechanism first analyzes the digital sandbox deduction logs and data blood sources to accurately locate the core high-level functional modules that cause the systematic failure of the solution, and disassembles the failed high-level components according to the predefined atomic rule library, deconstructing them into a set of fingerprint high-level components with clear input and output interfaces and independent functions, with finer granularity. For example, "pollution diffusion prediction" is disassembled into basic steps such as "buffer generation", "flow direction simulation", "concentration calculation", and "thermal map rendering".

[0121] Furthermore, during the disassembly process, the tight coupling dependencies of the original components are released to ensure that each fingerprint high-order component represents a reusable business atomic operation unit with clear functional semantics, and a unique topological embedding vector is generated for it to identify its functional semantics and interface characteristics.

[0122] Furthermore, the digital sandbox environment builds a digital twin that can simulate the complexity of reality and dynamically inject multi-level perturbations:

[0123] Specifically, multi-level disturbances include data layer anomalies, device layer failures, and business layer conflicts.

[0124] Furthermore, data layer anomalies include real defects such as simulated positioning drift and data loss; device layer failures include reproducing hardware bottlenecks such as computing node delays and memory overflow.

[0125] Business layer conflicts include logical contradictions such as mutually exclusive construction rules and excessive authority.

[0126] In the present application, a preferred set of solutions that achieves an optimal balance in multiple dimensions may include robustness, efficiency, cost, and user adaptability.

[0127] Specifically, robustness is used to measure the stability of the solution in failure scenarios; efficiency is used to evaluate end-to-end processing latency and resource consumption; cost is used to quantify computing resources and bandwidth overhead; and user adaptability is used to analyze the fit between operation paths and historical behavior patterns.

[0128] In the present application, a preferred method of screening invalid solutions that are significantly inferior to other solutions in all evaluation dimensions can be to exclude the solutions that are comprehensively degraded and retain the optimal solution set with balanced multi-dimensional indicators as the final decision.

[0129] The fingerprint high-order components are freely combined, and the high-order components are abstracted into topological embedding vectors, so that components with similar functions are adjacent to the topological embedding vectors in the vector space.

[0130] The functional semantics, topological associations and industry attributes of fingerprint high-order components are fused in multiple dimensions through the feature extraction engine, and the topological embedding model is trained based on the contrastive learning framework.

[0131] The topological embedding model maps components with similar functional semantics and topological roles to adjacent positions in the vector space.

[0132] For example, based on the spatial expansion of the distance between locations, the spatial buffer analysis function is performed to generate the Euclidean distance of the topological embedding vector in the vector space. The calculated Euclidean distance is used to determine whether the irrelevant functions meet the user needs. If the Euclidean distance is less than the distance of the irrelevant functional components, the topology that meets the user needs is output. If the Euclidean distance is not less than the distance of the irrelevant functional components, it does not meet the user needs.

[0133] A virtual model is established to simulate the combination of fingerprint high-order components and inject virtual random disturbances, and the operation results of the high-order components are output. If the operation results fail, the simplification of the high-order components is triggered, and the constraints are adjusted to repair the failure of the results.

[0134] In the present application, a preferred specific implementation method of simplifying high-order components includes:

[0135] When the virtual model detects that the operating results are invalid, the system triggers a cascade simplification process.

[0136] S1 uses data lineage analysis technology to trace back abnormal data flows, accurately mark high-order component nodes that cause systematic deviations, and rely on sandbox logs to build a fault propagation tree to quantify the contribution of sub-module failures.

[0137] S2 calls a predefined atomic rule library to decouple faulty components into a chain of fine-grained basic operation units, while imposing triple simplification restrictions. Multiple simplification restrictions can include depth constraints, core protection, and interface inheritance.

[0138] S3 expands the tolerance range according to the gradient for failures caused by overly strict constraints; it can recalibrate parameters through the PID controller based on historical sandbox data; it extracts cross-domain repair cases from the federal knowledge base and generates adaptation strategies to inject new constraints.

[0139] S4 only performs sandbox retesting on simplified sub-chains, generates vectors for new sub-units through topological embedding vector updates and associates functionally similar components, shortens the Euclidean distance in the vector space to achieve semantic clustering, abstracts the simplified path into a degraded rule template and maps it into a federated knowledge base, forming a self-optimizing system for failure location, disassembly and repair, vector reorganization, and rule evolution.

[0140] The virtual model loads fingerprint high-order components and their topological embedding vectors through an adapter, and generates initial connection logic based on functional semantic similarity.

[0141] The virtual model also predefines the data layer, device layer and business layer through the perturbation rule library, and eliminates interference data in the database, device layer and business layer through multi-level perturbations.

[0142] A topological semantic matching algorithm automatically concatenates adjacent fingerprint high-order components in the vector space. For example, buffer analysis is mapped to heat map rendering, forming a complete workflow and synchronously activating multi-level perturbation injection.

[0143] The specific multi-level layers include data layer, device layer and business layer; the data layer injects simulated sensor errors such as coordinate drift and attribute missing; the device layer triggers virtual hardware failures to reproduce edge bottlenecks; the business layer implants rule-based mutually exclusive events to test logical fault tolerance, which is used to simulate real-world complexity.

[0144] When the running results deviate from the safety tolerance or the process is interrupted, the data is traced back to the failed high-order component, and the disassembler is called to deconstruct the failed high-order component into fine-grained basic units. At the same time, the disassembly depth is limited, the decomposition of key components is shielded, and a degradation fallback plan is generated.

[0145] The topology embedding vector library is retested to see if it is constructed by dynamically adjusting constraints, including relaxing boundaries, parameter calibration, federation rule evolution, and incremental verification.

[0146] The relaxation bound is used for gradient extension failure constraint.

[0147] The parameter calibration adjusts input parameters based on feedback.

[0148] The federated rule evolution extracts similar failure strategies from a cross-domain knowledge base.

[0149] The incremental verification is used to retest whether the topology embedding vector library is constructed. If the construction is successful, the topology embedding vector library is updated to form a closed-loop optimization.

[0150] In this application, a preferred technical solution includes:

[0151] The initial connection logic principle of the virtual model establishes component connections through a two-dimensional matching mechanism: functional semantic matching compares the relative positions of component topological embedding vectors in vector space. When the core functional goals of two components are strongly semantically related, as manifested by a close distance in vector space, the system determines that they are composable.

[0152] Furthermore, a two-dimensional matching mechanism establishes component connections: interface contract compatibility verifies whether the input / output data types and structures of candidate components meet the upstream and downstream transmission requirements, and connection edges are only allowed to be established when the data contracts are consistent.

[0153] The disassembly operation relies on a predefined industry atomic operation library, which decomposes complex functions into indivisible standard operation units based on domain knowledge. The sub-units inherit the input / output conventions of the original components to ensure that they can still interact with upstream and downstream components after downgrading. Disassembly is prohibited for modules involving security or core logic to maintain their atomicity to ensure that the system's key capabilities are not degraded.

[0154] Databases can be migrated across domains by establishing federated rule learning.

[0155] Cross-domain policy migration is achieved through a three-layer adaptation mechanism.

[0156] Furthermore, scenario similarity matching is based on the topological embedding vector distance, which retrieves historical cases with similar functional semantics to the current failure scenario, extracts the applicable boundaries of historical strategies, verifies their constraint compatibility with the current scenario, and automatically inserts lightweight adaptation modules to resolve technical differences.

[0157] Incremental verification and vector library updates only perform sandbox testing on simplified sub-component chains, significantly reducing verification overhead. When a new sub-unit passes verification, the system dynamically adjusts the position of its topological embedding vector in space to bring it closer to a group of components with similar functions. The successful simplification path is abstracted into reusable degradation rules and stored in the federated knowledge base for global call.

[0158] like Figure 3As shown, the starting node on the left side of the S1 demand relationship topology construction diagram is marked as S1, which extracts spatial entities, relationships, and behaviors from real scenes, constructs the demand relationship topology, and extracts spatial entities and constructs the topology.

[0159] L1 is the topology input workflow orchestration code, S1 points to the connection line on the right, marked as L1, inputs the topology into the network analysis unit and orchestrates the workflow. Inputs the topology into the network analysis unit and orchestrates the workflow at the same time.

[0160] S2 generates candidate processes in parallel. In the middle area, branches such as S2 and the first candidate solution are marked. Multiple candidate processes are generated in parallel by the orchestrated workflow.

[0161] Prioritize L2 candidate solutions, mark the connecting line between S2 and S3 with L2, perform multi-dimensional scoring and sorting on the candidate solutions, and screen the optimal solution set.

[0162] S3 virtual model injects disturbances,The virtual simulation model node in the figure is labeled S3.,Disturbances such as data anomalies and equipment failures are injected into the digital sandbox to simulate the,real environment.

[0163] The L3 simulation result has two bidirectional branches. The two branch lines output by S3 are marked with L3 success and L4 failure. If the simulation is successful, the solution is output; if it fails, the component disassembly is triggered.

[0164] S4 high-level component disassembly and repair. The high-level component on the right side of the figure simplifies the node and marks it as S4. The failed workflow is disassembled into fingerprint high-level components, and the constraints are adjusted to repair the failure.

[0165] After L4 optimization, the solution is retested, S4 points to the loop line of S3, marked with L4, and the repaired solution re-enters the virtual simulation to form a closed-loop optimization.

[0166] Example 2

[0167] like Figure 2 As shown, a GIS application development system based on low code.

[0168] Extract modules, arrange and disassemble modules, combine modules and virtual models.

[0169] The extraction module extracts spatial entities, spatial relationships and spatial behaviors from real scenes, constructs demand relationship topology, and the network analysis unit choreographs the workflow.

[0170] The orchestration and decomposition module is used to generate multiple candidate processes in parallel from the orchestrated workflow, simulate and deduce multiple solutions, and decompose the orchestration workflow into fingerprint high-level components.

[0171] The combination module freely combines the fingerprint high-order components, abstracts the high-order components into topological embedding vectors, and makes components with similar functions adjacent to the topological embedding vectors in the vector space.

[0172] The virtual model includes a component topology adapter, a perturbation rule base and dynamic constraints.

[0173] like Figure 4 As shown, there are three types of initial solutions: the first candidate solution, the second candidate solution, and the third candidate solution, which represent workflows generated from different strategic paths.

[0174] Each candidate solution first enters the unranked solution state, and after being screened by rules, it becomes a ranked solution, completing the preliminary distinction between advantages and disadvantages.

[0175] All ranked solutions are summarized in the solution evaluation phase, and based on user needs, two branches emerge:

[0176] If the requirements are met, the final result is output directly and the process ends.

[0177] If the requirements are not met, proceed to the disassembly and virtual verification phase, and break the orchestrated workflow into high-level fingerprint components for further optimization.

[0178] The disassembled fingerprint high-level components are input into the virtual simulation model and injected with virtual random perturbations. The verification is divided into two branches.

[0179] The success of the virtual simulation shows that the component solution can still meet the needs under disturbances and output a solution that meets user needs.

[0180] The failure of virtual simulation triggers the simplification of high-level components and the invalidation of repair results. The component logic and parameters are adjusted to try to make the solution meet the standards. After that, the simulation verification may be re-entered to form an optimization closed loop.

[0181] It is important to note that the configuration and arrangement of the present application, as illustrated in various exemplary embodiments, are illustrative only. Although only two embodiments are described in detail in this disclosure, those reading this disclosure should readily understand that the dimensions, scales, structures, shapes, and proportions of various components, as well as parameter values ​​(e.g., temperature, pressure, etc.), mounting arrangements, use of materials, colors, and orientations, can be varied without materially departing from the subject matter described herein. For example, components shown as integrally formed may be comprised of multiple parts or components, the positions of components may be inverted or otherwise altered, and the nature, number, or position of discrete components may be modified or changed. Therefore, all such modifications are intended to be encompassed within the scope of this invention. The order or sequence of any process or method steps may be altered or reordered according to alternative embodiments. Any "means-plus-function" clause is intended to cover structures that perform the functions described herein, and not only structural equivalence but also structural equivalents. Other substitutions, modifications, changes, and omissions may be made in the design, operating conditions, and arrangement of the exemplary embodiments without departing from the scope of this invention. Therefore, the invention is not limited to the specific embodiments, but extends to various modifications that still fall within the scope of the appended claims.

[0182] Additionally, in order to provide a concise description of exemplary embodiments, all features of an actual embodiment may not be described (ie, those features that are not relevant to the best mode presently contemplated for carrying out the invention or those that are not relevant to implementing the invention).

[0183] It should be understood that in the development of any actual embodiment, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but for those of ordinary skill having the benefit of this disclosure, the development effort will be a routine task of design, fabrication, and production without undue experimentation.

[0184] 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. A low-code-based GIS application development method, characterized in that: include: Extract spatial entities, spatial relationships, and spatial behaviors from real-world scenarios, construct a demand relationship topology, input the demand relationship topology into a network analysis unit, and orchestrate workflows. The demand relationship topology is imported into graph calculation, with spatial entities and spatial behaviors as heterogeneous nodes, and spatial relationships and logical dependencies as weighted edges, to construct the demand relationship topology. The demand relationship topology is converted into a network model, and the node and edge information of the topology is serialized into a format compatible with the network analysis unit, and input into the network analysis unit. A triple association matrix is ​​constructed to convert the demand relationship topology into a computable feature. The triple association matrix includes an adjacency matrix, a metric matrix, and a direction matrix. The network analysis unit analyzes the relationship topology and dynamically inserts a mutex to eliminate spatial conflicts and resource conflicts, and groups the behavior nodes without dependencies into the same level. The nodes in the same level are partitioned by spatial position and output to generate a directed acyclic graph with level numbering. Converting each spatial behavior of the directed acyclic graph into an executable unit; The orchestrated workflow generates multiple candidate processes in parallel, simulates and deduces multiple solutions, and selects the optimal solution by prioritization. If multiple solutions do not meet the requirements, the orchestrated workflow is disassembled into fingerprint high-level components; Freely combine the fingerprint high-order components, abstract the high-order components into topological embedding vectors, and make components with similar functions adjacent to the topological embedding vectors in the vector space; A virtual model is established to simulate the combination of fingerprint high-order components and inject virtual random perturbations, and the operation results of the high-order components are output. If the operation results fail, the simplification of the high-order components is triggered, and the constraints are adjusted to repair the failure of the results.

2. The low-code GIS application development method according to claim 1, characterized in that: Extract multiple spatial entities from real-world scenes through multimodal interaction, obtain the longitude and latitude coordinates of the multiple spatial entities, and access the spatial database to assign spatial fingerprints to the spatial entities. The spatial fingerprints include type, location, and attributes. The spatial relationship includes a topological relationship, a metric relationship, and a directional relationship. The topological relationship between multiple spatial entities is determined by the longitude and latitude coordinates of the spatial entities and the attributes of the spatial fingerprint. The distances between the multiple spatial entities are calculated from the longitude and latitude coordinates of the multiple spatial entities to determine the metric relationship. The directional relationship between the multiple spatial entities is determined by the vector direction of the longitude and latitude coordinate sequence. The user operates the spatial entity to output an operation instruction, and the spatial behavior is extracted by parsing the operation instruction and binding the spatial fingerprint; A demand relationship topology is constructed by the spatial entities, spatial relationships and spatial behaviors.

3. The low-code GIS application development method according to claim 1, characterized in that: The generation of multiple candidate processes includes: based on the feature representation of the demand relationship topology, searching historical process templates through a case matching engine, and directly reusing templates that meet similarity requirements; if there are no matching templates, launching a generative adversarial mechanism to create new candidate workflows and verify their logical completeness; The candidate workflows are simulated and verified in a digital sandbox environment. By injecting multimodal disturbances such as simulated equipment failures and data anomalies, it is detected whether the output of the solution deviates from expectations. When the solution is marked as failed, the dynamic degradation mechanism is triggered, and the corresponding fingerprint high-level components are disassembled into basic operation units. The degradation processing rules are refined collaboratively through the federated knowledge base to update the knowledge base.

4. The low-code GIS application development method according to claim 1, characterized in that: The orchestrated workflow is a diversion engine, which generates multiple candidate processes in parallel. The diversion engine automatically selects and activates three types of generation strategies; The three types of generation strategies described above retrieve the historical process template library to reuse workflows that meet the topological similarity requirements; apply predefined business logic rule chain reasoning to generate standard processes that comply with industry specifications; and use the generative adversarial mechanism to dynamically innovate and generate new workflows whose spatial logic rationality is verified.

5. The low-code-based GIS application development method according to claim 4 is characterized in that: Conduct multi-dimensional simulations of multiple candidate solutions generated in parallel in a digital sandbox environment. The digital sandbox dynamically injects multi-level disturbances such as data anomalies, equipment failures, and service conflicts to simulate the complexity of a real environment. During the deduction process, the adaptive evaluation system collects data on the robustness, efficiency, cost, and user suitability of each solution in real time, generates a dynamic score, and filters out invalid solutions that are significantly inferior to other solutions in all evaluation dimensions. It retains the set of solutions that achieve the best balance in multiple dimensions. The system then sorts the optimal solution set based on preset priority rules and user-defined preferences, and outputs the optimal solution for priority execution. When all the deduced candidate solutions fail to meet the core requirements, the system triggers the workflow intelligent disassembly mechanism.

6. The low-code-based GIS application development method according to claim 1, characterized in that: The feature extraction engine performs multi-dimensional feature fusion on the functional semantics, topological associations, and industry attributes of fingerprint high-level components, and trains a topological embedding model based on a contrastive learning framework. The topological embedding model maps components with similar functional semantics and topological roles to adjacent positions in the vector space.

7. The low-code-based GIS application development method according to claim 6, characterized in that: The virtual model loads fingerprint high-order components and their topological embedding vectors through an adapter, and generates initial connection logic based on functional semantic similarity; The virtual model also predefines the data layer, device layer, and business layer through a perturbation rule library, eliminating interference data in the database, device layer, and business layer through multi-level perturbations; The adjacent fingerprint high-order components in the vector space are automatically concatenated through a topological semantic matching algorithm.

8. The low-code-based GIS application development method according to claim 7, characterized in that: When the running results deviate from the safety tolerance or the process is interrupted, the data is traced back to the failed high-level component, and the disassembler is called to deconstruct the failed high-level component into fine-grained basic units. At the same time, the disassembly depth is limited, the decomposition of key components is blocked, and a degradation fallback plan is generated; Retesting whether the topological embedding vector library is constructed by dynamically adjusting constraints, wherein the dynamic adjustment of constraints includes relaxing boundaries, parameter calibration, federation rule evolution, and incremental verification; The relaxation bound is used for gradient extension failure constraint; The parameter calibration adjusts input parameters based on feedback; The federated rule evolution extracts similar failure strategies from a cross-domain knowledge base; The incremental verification is used to retest whether the topology embedding vector library is constructed. If the construction is successful, the topology embedding vector library is updated to form a closed-loop optimization.

9. A low-code-based GIS application development system, which is implemented based on a low-code-based GIS application development method according to any one of claims 1 to 8, characterized in that: The system includes: an extraction module, an arrangement and disassembly module, a combination module and a virtual model; The extraction module extracts spatial entities, spatial relations and spatial behaviors from real scenes and constructs demand relationship topology, and the network analysis unit choreographs the workflow; The demand relationship topology is converted into a network model. The node and edge information of the topology is serialized into a format compatible with the network analysis unit and input into the network analysis unit. The demand relationship topology is converted into computable features by constructing a triple correlation matrix consisting of an adjacency matrix, a metric matrix, and a direction matrix. The orchestration and disassembly module is used to generate multiple candidate processes in parallel from the orchestrated workflow, simulate and deduce multiple solutions, and disassemble the orchestration workflow into fingerprint high-level components; The combination module freely combines the fingerprint high-order components, abstracts the high-order components into topological embedding vectors, and makes components with similar functions adjacent to the topological embedding vectors in the vector space; The virtual model includes a component topology adapter, a perturbation rule base and dynamic constraints.

Images