Space intelligent path generation method and device based on space function topology and medium

Abstract

The application provides a space intelligent path generation method and device based on a space function topology and a medium, belongs to the technical field of space intelligence, and can make space intelligent navigation and path planning technology not only depend on geometric distance, but also model space function semantics and topological logical constraints. The space intelligent path generation method based on the space function topology comprises the following steps: acquiring structural information of a space environment to construct a topological graph, wherein the structural information comprises space structural information and function area information; performing information propagation and aggregation on nodes of the topological graph by using a graph neural network to generate a global topological embedding matrix; and selecting an optimal path between nodes based on the global topological embedding matrix to make the path reachable in a physical space and meet logical constraints. The application exemplarily discloses a space intelligent path generation device based on a space function topology and a computer storage medium.

Description

Technical Field

[0001] This invention relates to the field of spatial intelligence technology, and in particular to a spatial intelligent path generation method, device and medium based on spatial functional topology. Background Technology

[0002] Currently, path planning and reasoning techniques in spatial intelligent systems are primarily based on geometric spatial representation and low-level accessibility search. Common approaches include discrete path search methods based on grid maps or voxels, shortest path planning based on graph theory, and end-to-end navigation networks based on deep learning. These methods have been widely researched and applied in areas such as robot navigation, autonomous driving, intelligent security, and augmented reality path prompting. However, existing technologies generally rely on geometric spatial information for path reasoning and generation, and their spatial cognition and reasoning capabilities are still mainly concentrated at the "geometric accessibility" level. They lack modeling and reasoning mechanisms for functional semantic relationships, behavioral constraint logic, and topological consistency in space, resulting in significant limitations in complex and multifunctional spatial environments.

[0003] This limitation is mainly manifested in the following aspects: (1) The spatial functional relationship lacks topological expression and structural constraints. Traditional path generation algorithms are usually based on distance metrics or cost function optimization in coordinate space, with the shortest distance or lowest energy consumption as the main goal. However, in real building, park or industrial environments, space is not only composed of geometric structures, but also contains complex functional divisions (such as office areas, rest areas, passage areas, restricted areas, etc.). Existing methods are difficult to express the logical constraint relationship between these functional areas, such as semantic rules such as "rest areas should be adjacent to office areas", "passages are one-way", and "restricted areas are prohibited from crossing", which leads to the generated path being geometrically feasible, but may be unreasonable or unsafe in functional logic. (2) The spatial functional relationship lacks topological expression and structural constraints. Some studies have attempted to introduce semantic labels of nodes and edges in semantic maps or scene graphs to describe the relationship between spatial entities. However, such methods often remain at the weak association at the label level and lack systematic modeling of the functional topology between functional nodes. Current path reasoning networks only accept local geometric features or image semantic embeddings at the input layer, failing to encode the topological connection patterns of spatial functions into a learnable embedding space, making it difficult to achieve global path reasoning based on functional relationships. Furthermore, there are issues such as the separation of semantic and geometric reasoning processes, the lack of a unified modeling mechanism, the lack of semantic consistency constraints in path generation, the lack of adaptive and robust learning mechanisms for functional topological changes, and the absence of a hierarchical reasoning structure, making it difficult to balance global rationality and local optimality… This disclosure aims to seek solutions to these problems. Summary of the Invention

[0004] This invention provides a spatial intelligent path generation method, device, and medium based on spatial functional topology, which can solve the problem that existing spatial intelligent navigation and path planning technologies rely only on geometric distance and lack modeling of spatial functional semantics and topological logical constraints.

[0005] In a first aspect, a spatial intelligent path generation method based on spatial functional topology is provided, comprising: acquiring structural information of the spatial environment to construct a topology graph, wherein the structural information includes spatial structure information and functional area information; using a graph neural network to perform information propagation and aggregation on the nodes of the topology graph to generate a global topology embedding matrix; and selecting the optimal path between nodes based on the global topology embedding matrix, so that the path is reachable in physical space and satisfies logical constraints.

[0006] In some embodiments, acquiring structural information in a spatial environment to construct a topology graph includes the following steps: collecting spatial structural information of the spatial environment; dividing the spatial environment into functional regions using semantic segmentation and region recognition algorithms, and collecting functional region information; using each functional region as a topology node, establishing topology edges between two topology nodes that have a reachability relationship or a logical relationship, wherein the topology edges satisfy the mathematical relation:

[0007]

[0008] in, Indicates the connection weights between topological nodes; Represents the topological node corresponding to the functional area of ​​the spatial environment; i and j are positive integer indices; Indicates the predetermined topology node The geometric distance function between them; Indicates the predetermined topology node The semantic similarity function is used to measure the semantic logical correlation between functional areas; Indicates the predetermined topology node The probability of transition between different functional areas; , These represent the predetermined weighting coefficients for geometric distance, semantic logical relevance, and transition probability, respectively.

[0009] In some embodiments, a graph neural network is used to propagate and aggregate information about the nodes of the topology graph to generate a global topology embedding matrix, including the following steps: obtaining all node information of the topology graph, the node information including function labels, spatial coordinates, and topological edge information; updating the node information, satisfying the formula:

[0010]

[0011] in, Represents topology nodes Features in the (l+1)th layer; i, j are positive integer indices; Represents topology nodes Features in the l-th layer; Represents topology nodes Features in the l-th layer; Represents topology nodes The set of adjacent nodes; Represents topology nodes and The topological edge feature values ​​between them include functional label feature values, spatial coordinate feature values, and topological edge information feature values; This represents a learnable weight matrix used for linear transformation of the self-features of topological nodes; This represents a learnable weight matrix used for linear transformation of neighbor information of topological nodes; Represents the predetermined update iteration coefficients; generates the global topological embedding matrix. Where H represents the global topological embedding matrix; This represents the information of the nth node.

[0012] In some embodiments, selecting the optimal path between nodes based on the global topology embedding matrix, such that the path is physically reachable and satisfies logical constraints, includes the following steps: based on the global topology embedding matrix, using a neural network architecture based on deep sequence modeling, and combining the feature values ​​of the spatial structure information and functional region information, selecting nodes that are physically reachable; during the inference process, evaluating whether each candidate path satisfies the logical constraints, and if so, determining it as a selectable path, and obtaining the optimal path from the selectable paths through iteration.

[0013] In some embodiments, the method further includes: adding interference nodes for data perturbation to the structure of the topology graph to obtain at least one perturbation path; determining whether the at least one perturbation path satisfies the physical reachability and logical constraints, and if not, adjusting the corresponding parameters.

[0014] In some embodiments, the method further includes: generating functional region information sequences based on the node information of the topology graph; and updating the learnable weight matrix based on the functional region information sequences. and learnable weight matrix .

[0015] Secondly, a spatial intelligent path generation device based on spatial functional topology includes: a spatial functional topology construction module, used to acquire structural information of the spatial environment to construct a topology graph, wherein the structural information includes spatial structure information and functional area information; a functional topology embedding module, used to use a graph neural network to perform information propagation and aggregation on the nodes of the topology graph to generate a global topology embedding matrix; and a path generation module, used to select the optimal path between nodes based on the global topology embedding matrix, so that the path is reachable in physical space and satisfies logical constraints.

[0016] In some embodiments, the system further includes: a topology self-supervised optimization module, configured to add interference nodes for data perturbation to the structure of the topology graph to obtain at least one perturbation path; determine whether the at least one perturbation path satisfies the reachability and logical constraints in the physical space, and if not, adjust the corresponding parameters.

[0017] Thirdly, an electronic device is provided, comprising: at least one processor; and at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to utilize the at least one processor to cause the electronic device to perform the above-described spatial intelligent path generation method based on spatial functional topology.

[0018] Fourthly, a computer-readable storage medium is also provided, wherein program code is stored in the computer-readable storage medium, and the program code can be called by a processor to execute the above-described spatial intelligent path generation method based on spatial functional topology.

[0019] Compared with the prior art, the present invention can achieve at least the following beneficial effects: it enables spatial intelligent navigation and path planning technology to not only rely on geometric distance, but also to model spatial functional semantics and topological logical constraints.

[0020] The summary section is provided to present the chosen concepts in a simplified form, which will be further described in the detailed description below. The summary section is not intended to identify essential or indispensable features of this disclosure, nor is it intended to limit the scope of this disclosure. Attached Figure Description

[0021] The above and other objects, features and advantages of this disclosure will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.

[0022] Figure 1 A schematic diagram of a spatial intelligent path generation method based on spatial functional topology provided in an embodiment of this application is shown;

[0023] Figure 2 A schematic diagram of the flow steps of a spatial intelligent path generation method based on spatial functional topology provided in one embodiment of this application is shown;

[0024] Figure 3 This illustration shows a functional topology embedding process provided in yet another embodiment of this application.

[0025] Figure 4 A schematic diagram of a spatial intelligent path generation device based on spatial functional topology provided in an embodiment of this application is shown.

[0026] Figure 5 A schematic diagram of an electronic device using a spatial intelligent path generation method based on spatial functional topology provided in an embodiment of this application is shown. Detailed Implementation

[0027] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0028] The term "comprising" and its variations as used herein signify open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "at least partially based on". The terms "one example embodiment" and "one embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0029] This application provides a spatial intelligent path generation method based on spatial functional topology. Please refer to [link / reference]. Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1 The first embodiment of this application provides a detailed description of a spatial intelligent path generation method 100 based on spatial functional topology.

[0030] Step S102: Obtain data information, that is, obtain the structural information of the spatial environment to construct a topology map, wherein the structural information includes spatial structure information and functional area information.

[0031] Step S104: Generate a topological embedding matrix, that is, use a graph neural network to perform information propagation and aggregation on the nodes of the topological graph to generate a global topological embedding matrix.

[0032] Step S106: Select a path, that is, based on the global topology embedding matrix, select the optimal path between nodes so that it is reachable in physical space and satisfies logical constraints.

[0033] In some embodiments, acquiring structural information in a spatial environment to construct a topology graph includes the following steps: collecting spatial structural information of the spatial environment; dividing the spatial environment into functional regions using semantic segmentation and region recognition algorithms, and collecting functional region information; using each functional region as a topology node, establishing topology edges between two topology nodes that have a reachability relationship or a logical relationship, wherein the topology edges satisfy the mathematical relation:

[0034]

[0035] in, Indicates the connection weights between topological nodes; Represents the topological node corresponding to the functional area of ​​the spatial environment; i and j are positive integer indices; Indicates the predetermined topology node The geometric distance function between them; Indicates the predetermined topology node The semantic similarity function is used to measure the semantic logical correlation between functional areas; Indicates the predetermined topology node The probability of transition between different functional areas; , These represent the predetermined weighting coefficients for geometric distance, semantic logical relevance, and transition probability, respectively.

[0036] For example: Input and data acquisition: The system first collects environmental data from spatial perception terminals, including spatial structure information from LiDAR, cameras, BIM models, semantic maps, or SLAM systems. Through semantic segmentation and region recognition algorithms, the space is divided into functional areas, such as office areas, corridors, meeting areas, rest areas, equipment rooms, or restricted areas.

[0037] Functional Node and Topology Edge Definition: Each functional region is defined as a topology node. Node attributes include: function type; geometric center coordinates and boundary description; traffic direction and access conditions; area usage constraints (such as one-way access only or access by specific roles). If there is a reachability relationship or logical connection between two functional areas (such as passage connection, hierarchical adjacency, or functional dependency), a topological edge is established between the two nodes. The edge weight can be determined by the following factors:

[0038]

[0039] To facilitate understanding of the technical implementation mechanism of this invention, the key parameters in the invention and their definitions, mathematical meanings, and functional relationships in the spatial functional topology path reasoning system are explained as follows.

[0040] in: It indicates functional nodes in the space, such as office areas, corridors, and rest areas; The connection weight between nodes is used to measure the passage cost or logical connection strength between two functional areas; It represents a geometric distance function, reflecting the physical accessibility of two functional areas in space; This represents a semantic similarity function that measures the semantic and logical correlation between functional areas. The probability of functional transfer is represented by α1, α2, and α3, reflecting the logical rationality of transitioning from one functional area to another. α1, α2, and α3 are the weighting coefficients of the corresponding factors, used to balance the importance of geometry, semantics, and functional logic. Through this weighted model, this invention achieves a unified modeling of geometric reachability, semantic consistency, and functional logical rationality in path reasoning. This formula is the core mathematical foundation for the entire spatial functional topology construction and path reasoning, enabling the system to consider spatial geometric optimality, semantic logical correctness, and environmental adaptability during path generation, thereby achieving intelligent and interpretable spatial path planning. 3. Topology Construction and Output: The module ultimately outputs a functional topology graph G=(V,E), where V is the set of functional nodes and E is the set of functional connections. The topology reflects both the spatial layout and the semantic logical relationships. 4. Role in the Overall Process: The spatial functional topology construction module, as the first step of the system, provides a structured semantic foundation for subsequent embedding and path generation, enabling the system to understand space from a functional perspective, rather than solely based on path reachability judgments based on geometric information.

[0041] In some embodiments, a graph neural network is used to propagate and aggregate information about the nodes of the topology graph to generate a global topology embedding matrix, including the following steps: obtaining all node information of the topology graph, the node information including function labels, spatial coordinates, and topological edge information; updating the node information, satisfying the formula:

[0042]

[0043] in, Represents topology nodes Features in the (l+1)th layer; i, j are positive integer indices; Represents topology nodes Features in the l-th layer; Represents topology nodes Features in the l-th layer; Represents topology nodes The set of adjacent nodes; Represents topology nodes and The topological edge feature values ​​between them include functional label feature values, spatial coordinate feature values, and topological edge information feature values; This represents a learnable weight matrix used for linear transformation of the self-features of topological nodes; This represents a learnable weight matrix used for linear transformation of neighbor information of topological nodes; Represents the predetermined update iteration coefficients; generates the global topological embedding matrix. Where H represents the global topological embedding matrix; This represents the information of the nth node.

[0044] For example, in the functional topology embedding layer, 1. Input and target: The input is the aforementioned functional topology graph G=(V,E). Each node... It includes function labels, spatial coordinates, and edge connection information. The goal of this module is to generate embedded representations of function nodes. To capture spatial semantics and structural relationships.

[0045] Embedding mechanism and model structure:

[0046] A graph neural network is used to propagate and aggregate information between functional nodes. The node update process is as follows:

[0047]

[0048] in: :node Feature representation at layer l;

[0049] :node The set of adjacent nodes;

[0050] :node and Edge characteristics between them (such as geometric distance, functional relationship, etc.);

[0051] , Learnable weight matrices are used for linear transformations of node self-features and neighbor information, respectively.

[0052] : The operation of concatenating adjacent nodes and edge features.

[0053] Each node retains its own semantic information during the update process and aggregates the functional features of its neighboring nodes and edges, thereby achieving joint learning of semantic logic and spatial topology. After multiple updates, the model obtains a global topological embedding vector, which is used for subsequent path reasoning and semantic consistency constraints. This method enables the system to understand spatial functional relationships structurally, providing a logically sound and semantically consistent basic feature representation for path generation.

[0054] Output result:

[0055] The model outputs a global topological embedding matrix. Each node's embedding vector contains both functional semantics and spatial connectivity information.

[0056] Its role in the implementation process:

[0057] This module transforms the spatial functional topology into a continuous high-dimensional semantic space, enabling the path reasoning model to possess functional logic-level perception capabilities. The embedding results are directly input into the path generation module to guide subsequent path logic reasoning.

[0058] In some embodiments, selecting the optimal path between nodes based on the global topology embedding matrix, such that the path is physically reachable and satisfies logical constraints, includes the following steps: based on the global topology embedding matrix, using a neural network architecture based on deep sequence modeling, and combining the feature values ​​of the spatial structure information and functional region information, selecting nodes that are physically reachable; during the inference process, evaluating whether each candidate path satisfies the logical constraints, and if so, determining it as a selectable path, and obtaining the optimal path from the selectable paths through iteration.

[0059] For example, this module generates path sequences that conform to semantic logic constraints within a functional topology embedding space. By performing semantic modeling and logical reasoning on the embedded representations of spatial functional nodes, this module can generate path results that satisfy both geometric reachability and spatial functional logical relationships.

[0060] Input and objective:

[0061] The module's input includes the starting node, the target node, and a set of functional topology semantic features generated by the functional topology embedding layer. The overall goal of the system is to infer an optimal path from the starting node to the target node within this embedding space. This path must not only be physically reachable but also comply with logical constraints between spatial functions, such as regional access rules, functional connection order, and semantic consistency.

[0062] Path generation mechanism:

[0063] The module employs a neural network architecture based on deep sequence modeling, generating path sequences through step-by-step reasoning. During path generation, the model combines spatial functional embedding features, semantic relationships between nodes, and historical path patterns to predict node selection at each step. During reasoning, the system evaluates the logical consistency of each candidate path, including determining whether the path violates functional logic constraints (e.g., traversing restricted areas, reversing unidirectional passages, crossing functionally discontinuous areas), and comprehensively analyzes the semantic differences of the paths. The model undergoes multiple iterative optimizations during training to ensure that the generated paths remain coherent and logically sound at the functional level.

[0064] Output and subsequent use:

[0065] The module outputs one or more path sequences that conform to functional logic constraints, along with semantic transition weight information between nodes. These outputs can be directly used as input to the lower-level geometry optimization network to further generate smooth and executable geometric paths.

[0066] Integration process and operation:

[0067] During system runtime, the module first calculates an initial candidate path set in the functional topology graph based on the input starting and target nodes. Then, the model utilizes the feature information embedded in the functional topology and semantic history patterns to logically filter and generate sequences from the candidate paths. During the training phase, the model automatically adjusts its internal parameters by continuously comparing the predicted paths with logically correct paths, thereby achieving optimization and self-learning of path logical consistency. During the inference phase, the system can flexibly adjust the weights of logical constraints according to different task scenarios, implementing various path generation strategies such as safety priority, efficiency priority, or semantic rationality priority.

[0068] In some embodiments, the method further includes: adding perturbation nodes to the structure of the topology graph to obtain at least one perturbation path; determining whether the at least one perturbation path satisfies the physical reachability and logical constraints, and adjusting the corresponding parameters if not. For example, during model training, perturbation is actively applied to the spatial functional topology, enabling the model to maintain correct path reasoning logic under incomplete or changed topological conditions, thereby forming a learning mechanism with self-healing characteristics. During the model training phase, the system first generates a copy with random perturbation based on the original functional topology graph to simulate changes or anomalies in the spatial structure in the real environment. Subsequently, the model performs path reasoning on the perturbated topology to obtain a new path prediction result. The system compares and analyzes this predicted path with the original standard path in terms of semantic structure, node order, and logical consistency, and adjusts the model's internal parameters through difference feedback. In this process, the model gradually learns how to recover correct path logic in the case of missing information, broken connections, or structural shifts, thereby forming a "self-supervised learning" capability. The model parameters are updated without manual annotation, but rather through self-comparison and self-correction. This mechanism significantly improves the model's adaptability and generalization ability to dynamic topological environments. The topology self-supervised optimization mechanism is executed periodically during the system training phase, forming a closed-loop self-learning system together with the semantically consistent path generation module and the functional topology embedding layer. When local changes occur in the spatial environment (such as path closure, node failure, or functional region adjustment), this mechanism can guide the model to relearn and correct the topological representation without manual intervention, ensuring that the generated paths maintain functional logic correctness and semantic coherence.

[0069] In some embodiments, the method further includes: generating functional region information sequences based on the node information of the topology graph; and updating the learnable weight matrix based on the functional region information sequences. and learnable weight matrix For example, implementing collaborative reasoning between semantic logic planning and geometric path optimization. 1. Architecture Composition: Upper-layer semantic planning network: takes functional topology embedding as input, responsible for global functional path reasoning, determining the passage sequence of functional regions; Lower-layer geometric optimization network: performs local path smoothing, obstacle avoidance, and distance optimization based on a spatial geometric map. Two-layer collaborative mechanism: the functional region sequence output by the upper layer serves as the constraint input for the lower-layer geometric optimization; the geometric reachability calculated by the lower layer is fed back to the upper-layer network to update the functional path weights, achieving semantic-geometric bidirectional coupling. Implementation steps: input the functional path sequence output by the semantic layer into the geometric layer; the geometric layer performs local optimization on the physical map to generate the actual driving path; the optimization result is then fed back to the semantic layer for joint loss update; finally, a path scheme satisfying both functional logic and geometric constraints is output.

[0070] Figure 2 Another embodiment of this disclosure is shown. Figure 2 The diagram shows: S201 represents the fusion of LiDAR, depth camera, and BIM model data, followed by extrinsic parameter calibration and ICP algorithm to unify the coordinate system, generating a semantically parsable 3D environment model; S202 represents the automatic division of spatial functional units based on semantic segmentation algorithm, extracting key attributes such as functional labels, geometric centers, boundary shapes, and traffic directions; S203 represents the use of KD-Tree to calculate regional adjacency relationships, and the generation of weighted edges by integrating geometric distance, semantic similarity, and logical constraints to establish functional node associations; S204 represents the use of NetworkX to construct a directed functional topology graph, with node and boundary attributes stored in a sparse matrix form in an SQLite database, forming a structured topology model.

[0071] Figure 3 Another embodiment of this disclosure is shown. Figure 3 The diagram illustrates the following: S301 indicates that the system introduces a multimodal fusion strategy during the feature construction stage, jointly encoding the spatial geometry, functional labels, and edge weight correlation of nodes into high-dimensional sparse vectors. During normalization, feature denoising and distribution stretching are performed simultaneously, ensuring that different feature dimensions have statistically consistent contribution weights, thereby improving the convergence stability of embedding learning. S302 indicates that the improved GraphSAGE framework introduces an attention-weighted aggregation mechanism, allowing nodes to dynamically adjust the importance of different features during domain information propagation. During the training phase, path structure consistency constraints are employed, minimizing the distribution difference between real and predicted paths in the latent space through peer learning, thus achieving a joint expression of functional logic and spatial structure. S303 indicates that after the embedding stage, the system automatically evaluates the clustering consistency and embedding stability of node representations, filtering out abnormal node vectors to optimize model generalization performance. The final output of 128 semantic embedding vectors not only preserves the local structure of the functional topology but also possesses cross-scene transfer capabilities, and is encapsulated in HDF5 format for subsequent inference calls.

[0072] This application also provides a spatial intelligent path generation device 400 based on spatial functional topology. Please see below. Figure 4 The processing flow of the spatial intelligent path generation device 400 based on spatial functional topology may include the following modules: spatial functional topology construction module 402, functional topology embedding module 404, and path generation module 406.

[0073] The spatial functional topology construction module 402 is used to acquire structural information of the spatial environment to construct a topology map, wherein the structural information includes spatial structure information and functional area information;

[0074] The functional topology embedding module 404 is used to perform information propagation and aggregation on the nodes of the topology graph using a graph neural network to generate a global topology embedding matrix;

[0075] The path generation module 406 is used to select the optimal path between nodes based on the global topology embedding matrix, so that the path is reachable in physical space and satisfies logical constraints.

[0076] In some embodiments, the spatial intelligent path generation device 400 based on spatial functional topology may further include: a topology self-supervised optimization module, used to add interference nodes for data perturbation to the structure of the topology graph to obtain at least one perturbation path; and to determine whether the at least one perturbation path satisfies the reachability and logical constraints in physical space, and if not, to adjust the corresponding parameters.

[0077] like Figure 5 As shown in the illustration, an electronic device provided in this embodiment of the invention may include a processor 520 and a memory 510. Optionally, the electronic device may further include a transceiver. The processor, memory, and transceiver may be connected via a communication bus. The memory stores computer-readable instructions, which, when executed by the processor, implement the steps of the spatial intelligent path generation method based on spatial functional topology as described above.

[0078] In a specific implementation, as one example, processor 520 may include one or more CPUs.

[0079] Optionally, in a specific implementation, if the memory 510, processor 520, and communication interface 530 are integrated on a single chip, then the memory 510, processor 520, and communication interface 530 can communicate with each other through an internal interface.

[0080] In a specific implementation, as one example, the electronic device may also include multiple processors, each of which may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). Here, a processor may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0081] The memory is used to store the software program that executes the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can be referred to the above method embodiment, which will not be repeated here.

[0082] A transceiver is used to communicate with network devices or with terminal devices.

[0083] Optionally, the transceiver may include a receiver and a transmitter. The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0084] Optionally, the transceiver can be integrated with the processor or exist independently and coupled to the processor through the interface circuit of the electronic device. This embodiment of the invention does not specifically limit this.

[0085] It should be noted that the structure of the electronic device described above does not constitute a limitation on the electronic device. Actual electronic devices may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. Furthermore, the technical effects of the electronic device can be referred to the technical effects of the above method embodiments, and will not be repeated here.

[0086] In an exemplary embodiment, the present invention also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the steps of the spatial intelligent path generation method based on spatial functional topology described above. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0087] This invention also provides an electronic device, which includes: a processor; and a memory storing computer-readable instructions. When the computer-readable instructions are executed by the processor, the above-described spatial intelligent path generation method based on spatial functional topology is implemented.

[0088] This invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores program code, which can be called by a processor to execute the above-described spatial intelligent path generation method based on spatial functional topology.

[0089] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0090] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0091] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0092] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0094] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0095] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A spatial intelligent path generation method based on spatial functional topology, characterized in that, include: Obtaining structural information about the space environment to construct a topology map includes the following steps: Collect spatial structure information of the space environment; The spatial environment is divided into functional areas using semantic segmentation and region recognition algorithms, and functional area information is collected. Each functional area is treated as a topological node. Topological edges are established between two topological nodes that have a reachability or logical relationship, and these topological edges satisfy the mathematical formula: in, Indicates the connection weights between topological nodes; Represents the topological node corresponding to the functional area of ​​the spatial environment; i and j are positive integer indices; Indicates the predetermined topology node The geometric distance function between them; Indicates the predetermined topology node The semantic similarity function is used to measure the semantic logical correlation between functional areas; Indicates the predetermined topology node The probability of transition between different functional areas; , The predetermined weighting coefficients represent geometric distance, semantic logical relevance, and transition probability, respectively, wherein the structural information includes spatial structural information and functional region information; The process of using a graph neural network to propagate and aggregate information from the nodes of the topology graph to generate a global topology embedding matrix includes the following steps: Obtain all node information of the topology graph, including function labels, spatial coordinates, and topological edge information; Update the node information, and satisfy the formula: in, Represents topology nodes Features in the (l+1)th layer; i, j are positive integer indices; Represents topology nodes Features in the l-th layer; Represents topology nodes Features in the l-th layer; Represents topology nodes The set of adjacent nodes; Represents topology nodes and The topological edge feature values ​​between them include functional label feature values, spatial coordinate feature values, and topological edge information feature values; This represents a learnable weight matrix used for linear transformation of the self-features of topological nodes; This represents a learnable weight matrix used for linear transformation of neighbor information of topological nodes; Indicates the predetermined update iteration coefficients; Generate a global topological embedding matrix Where H represents the global topological embedding matrix; This represents the information of the nth node; Based on the global topology embedding matrix, the optimal path between nodes is selected so that it is reachable in physical space and satisfies logical constraints.

2. The spatial intelligent path generation method based on spatial functional topology according to claim 1, characterized in that, Based on the global topology embedding matrix, the optimal path between nodes is selected so that it is physically reachable and satisfies logical constraints, including the following steps: Based on the global topological embedding matrix, using a neural network architecture based on deep sequence modeling, and combining the feature values ​​of the spatial structure information and functional region information, nodes reachable by the path in the physical space are selected. During the reasoning process, each candidate path is evaluated to see if it meets the logical constraints. If it does, it is determined to be an optional path. The optimal path is obtained from the optional paths through iteration.

3. The spatial intelligent path generation method based on spatial functional topology according to any one of claims 1 to 2, characterized in that, Also includes: Add interference nodes for data perturbation to the structure of the topology graph to obtain at least one perturbation path; Determine whether the at least one perturbation path satisfies the physical reachability and logical constraints. If not, adjust the corresponding parameters.

4. The spatial intelligent path generation method based on spatial functional topology according to claim 1, characterized in that, Also includes: Based on the node information of the topology graph, functional area information sequences are generated respectively; Update the learnable weight matrix based on the functional area information sequence. and learnable weight matrix .

5. A spatial intelligent path generation device based on spatial functional topology, characterized in that, include: The spatial functional topology construction module is used to acquire structural information about the spatial environment and construct a topology map, including the following steps: Collect spatial structure information of the space environment; The spatial environment is divided into functional areas using semantic segmentation and region recognition algorithms, and functional area information is collected. Each functional area is treated as a topological node. Topological edges are established between two topological nodes that have a reachability or logical relationship, and these topological edges satisfy the mathematical formula: in, Indicates the connection weights between topological nodes; Represents the topological node corresponding to the functional area of ​​the spatial environment; i and j are positive integer indices; Indicates the predetermined topology node The geometric distance function between them; Indicates the predetermined topology node The semantic similarity function is used to measure the semantic logical correlation between functional areas; Indicates the predetermined topology node The probability of transition between different functional areas; , The predetermined weighting coefficients represent geometric distance, semantic logical relevance, and transition probability, respectively, wherein the structural information includes spatial structural information and functional region information; The functional topology embedding module is used to perform information propagation and aggregation on the nodes of the topology graph using a graph neural network to generate a global topology embedding matrix, including the following steps: Obtain all node information of the topology graph, including function labels, spatial coordinates, and topological edge information; Update the node information, and satisfy the formula: in, Represents topology nodes Features in the (l+1)th layer; i, j are positive integer indices; Represents topology nodes Features in the l-th layer; Represents topology nodes Features in the l-th layer; Represents topology nodes The set of adjacent nodes; Represents topology nodes and The topological edge feature values ​​between them include functional label feature values, spatial coordinate feature values, and topological edge information feature values; This represents a learnable weight matrix used for linear transformation of the self-features of topological nodes; This represents a learnable weight matrix used for linear transformation of neighbor information of topological nodes; Indicates the predetermined update iteration coefficients; Generate a global topological embedding matrix Where H represents the global topological embedding matrix; This represents the information of the nth node; The path generation module is used to select the optimal path between nodes based on the global topology embedding matrix, so that the path is reachable in physical space and satisfies logical constraints.

6. The spatial intelligent path generation device based on spatial functional topology according to claim 5, characterized in that, Also includes: The topology self-supervised optimization module is used to add interference nodes for data perturbation to the structure of the topology graph to obtain at least one perturbation path; Determine whether the at least one perturbation path satisfies the physical reachability and logical constraints. If not, adjust the corresponding parameters.

7. An electronic device, characterized in that, include: At least one processor; as well as At least one memory including computer program code, the at least one memory and the computer program code being configured to utilize the at least one processor to cause the electronic device to perform the spatial intelligent path generation method based on spatial functional topology as described in any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code, which can be called by a processor to execute the spatial intelligent path generation method based on spatial functional topology as described in any one of claims 1 to 4.

Images