Method, server and system for performing global path planning using building information modeling and physics engine

Abstract

The present application provides a kind of by global path planning server, utilizes building information modeling (BIM) and physical engine to execute the computer-implemented method for global path planning of unmanned ground vehicle (UGV) in indoor space. The method comprises: receiving the input modeling data of target UGV and designed task planning data;According to the input modeling data, carry out building information modeling (BIM) based information integration and map generation, so as to obtain the integrated BIM model of UGV and the semantic map of indoor space;According to the integrated BIM model of UGV, the semantic map, set target UGV center path planning algorithm, so as to generate the robot operating system (ROS) control data corresponding to target UGV;And according to the integrated BIM model of UGV, the semantic map, set target UGV center path planning algorithm and designed task planning data, carry out trajectory generation and task coordination based on physical engine, so as to obtain the final UGV path of the target UGV.

Description

Technical Field

[0001] This invention relates to path planning technology, and more specifically, to a method, server, and system for performing global path planning (GPP) of unmanned ground vehicles (UGVs) in indoor spaces using building information modeling (BIM) and a physical engine. Background Technology

[0002] Global construction-related spending amounts to approximately $10 trillion annually (equivalent to 13% of GDP), and this figure is projected to reach $14 trillion by 2025. While the construction industry is one of the largest sectors in the global economy, robotics has recently been recognized as a promising technology for improving productivity in building and facility management by automating time-intensive and labor-intensive tasks. Unmanned aerial vehicles (UGVs), in particular, are gaining more attention in indoor environments than other robots, such as drones, due to their larger payload capacity, longer battery life, and safer movement. Examples of these UGV agents include infrastructure inspection UGVs, indoor service UGVs, and construction UGVs. Among all the fundamental technical capabilities of UGVs, path planning is essential for autonomous navigation, and it consists of global path planning (GPP) and local path planning (LPP). Unlike LPP, which focuses only on a segment of the global path to avoid dynamic obstacles, GPP calculates the globally optimal path within the known environment, making it a higher-level and priority task for LPP.

[0003] Therefore, designing an efficient GPP system to improve the operating efficiency of UGVs in indoor environments is a goal that researchers in this field are dedicated to developing. Summary of the Invention

[0004] According to a first aspect of the present invention, a computer-based method is provided for performing global path planning of an unmanned ground vehicle (UGV) in an indoor space using a global path planning server, utilizing building information modeling (BIM) and a physics engine. The method includes: receiving input modeling data and designed task planning data of a target UGV; performing BIM-based information integration and map generation based on the input modeling data to obtain an integrated BIM model of the UGV and a semantic map of the indoor space; setting a target UGV central path planning algorithm based on the integrated BIM model and the semantic map to generate corresponding Robot Operating System (ROS) control data for the target UGV; and performing physics engine-based trajectory generation and task coordination based on the integrated BIM model, the semantic map, the set target UGV central path planning algorithm, and the designed task planning data to obtain a final UGV path for the target UGV, wherein the final UGV path and the ROS control data are sent to the target UGV.

[0005] According to a second aspect of the invention, a server is provided for performing global path planning for unmanned ground vehicles (UGVs) in indoor spaces using building information modeling and a physics engine. The server includes a processor for executing machine instructions to implement the methods described above.

[0006] According to a third aspect of the present invention, a system is provided for performing global path planning for an unmanned ground vehicle (UGV) in an indoor space using building information modeling and a physics engine. The system includes the server and the UGV. The server includes a processor for executing machine instructions to implement the method described above. Attached Figure Description

[0007] Figure 1 A block diagram of the provided global path planning system is shown.

[0008] Figure 2 A flowchart of the provided global path planning method is shown.

[0009] Figure 3 A schematic diagram of the provided global path planning system is shown.

[0010] Figure 4 A flowchart illustrating the provided global path planning method is shown.

[0011] Figure 5A It shows Figure 2 The flowchart for step S220.

[0012] Figure 5B It shows Figure 2The flowchart of step S230.

[0013] Figure 5C It shows Figure 2 The flowchart for step S240.

[0014] Figure 5D It shows Figure 2 The flowchart for step S244.

[0015] Figure 6 The diagram shows the architecture of the industrial base classification of UGV entities.

[0016] Figure 7A A schematic diagram of layered map generation is shown.

[0017] Figure 7B Another schematic diagram of layered map generation is shown.

[0018] Figure 8 Schematic diagrams of various mobility systems of UGV are shown.

[0019] Figure 9 A schematic diagram of the basic extension of UGV is shown.

[0020] Figure 10 A schematic diagram of the operation of a bicycle model of a restricted UGV is shown.

[0021] Figure 11A A schematic diagram showing the location information of UGVs in a grid map is provided.

[0022] Figure 11B The diagram illustrates both a failed case and a feasible case of UGV in a narrow corridor.

[0023] Figure 12A A schematic diagram of the provided three-stage inspection method is shown.

[0024] Figure 12B A flowchart of the provided three-stage inspection method is shown.

[0025] Figure 13A A schematic diagram is shown for the analysis extension for non-restrictive UGVs.

[0026] Figure 13B A schematic diagram of the analysis extension for restricted UGVs is shown.

[0027] Figure 14A A schematic diagram of a 2D geometric mesh path for a UGV is shown.

[0028] Figure 14B A schematic diagram of a 3D spatiotemporal blocky trajectory of a UGV is shown.

[0029] Figure 15A A schematic diagram of the 2D geometric mesh paths of multiple UGVs is shown.

[0030] Figure 15B A schematic diagram of the 3D spatiotemporal block trajectories of multiple UGVs is shown.

[0031] Figure 16 A schematic diagram of time slots for different grids is shown. Detailed Implementation

[0032] In the following description, a global path planning method, server, and system for unmanned ground vehicles (UGVs) in indoor spaces are used as preferred examples. It will be apparent to those skilled in the art that modifications, including additions and / or substitutions, can be made without departing from the scope and spirit of the invention. Specific details may be omitted to avoid obscuring the invention; however, this disclosure is prepared to enable those skilled in the art to practice the teachings herein without excessive experimentation. The general principles defined herein may be applied to other embodiments. Therefore, the invention is not intended to be limited to the specific embodiments shown and described, but should be given the broadest scope consistent with the principles and novel features disclosed herein.

[0033] Typically, there are three main tasks in UGV GPP: (1) map (environment) generation, (2) path planning algorithm, and (3) path collision avoidance. Nowadays, indoor environments present a significant challenge to GPP tasks due to their more complex structures (irregularly shaped spaces), various indoor restrictions, dynamic activities, and limited space. Specifically, complex environments directly increase the difficulty of map generation. For UGVs using Simultaneous Localization and Mapping (SLAM), sensor noise makes it difficult to accurately cover the environment, and the lack of semantic information can easily lead to violations of building restrictions. Some scholars have suggested using prior knowledge from sources such as SLAM, emergency maps, and BIM models to reduce the need for and cost of data collection. However, prior maps mainly consist of geometric information of building elements (such as walls and doors). A semantic map that comprehensively considers both real-world updates and UGV attributes is still needed. Furthermore, the limited space in indoor environments requires path planning algorithms to have sufficient accuracy. Therefore, the mobility and geometric characteristics of UGVs should be carefully integrated into each path planning component. Finally, trajectory (timestamped path) conflicts frequently occur in indoor environments, especially in crowded places (such as corridors). Previous studies have largely addressed trajectory conflicts based on the assumption that UGVs move at a constant or average speed. However, reality may be quite different.

[0034] Therefore, the following embodiments provide a trajectory generation and coordination method that simultaneously considers environmental and UGV attributes. Specifically, this invention proposes a comprehensive GPP system for multiple UGVs to operate simultaneously in an indoor environment. Three main improvements are proposed for the three main tasks of GPP: BIM-based information integration and map generation, a UGV-centric path planning algorithm, and trajectory generation and coordination based on a physics engine. A system is further designed. The specific objectives of this system are:

[0035] (1) Extending BIM applications to UGV map generation. Since map generation is the first task of GPP (Global Platform for Project Management), this embodiment aims to first establish a UGV-integrated BIM, enabling BIM and robot configuration to be integrated for better resource interoperability. This paper selects Industry Foundation Classes (IFC) as a representative BIM model. IFC is an open data standard based on Express, initiated by buildingSMART (formerly the International Interoperability Consortium) in 1994. IFC is a CAD data exchange model used to describe data related to buildings, architecture, and the construction industry. IFC is widely considered the most popular BIM standard, and most BIM software in the AEC (Architecture, Engineering, and Construction) industry supports the IFC standard. For UGV-integrated BIM, this system proposes a multi-layered map generation method to achieve clear map construction and efficient map maintenance.

[0036] (2) An improved indoor UGV path planning algorithm is proposed. As the second task of GPP, the path planning algorithm directly determines the path quality. To this end, this invention proposes an improved UGV center path planning algorithm to address the shortcomings of existing GPP algorithms. Four corresponding improvements are designed for UGV attributes: mobility-based primitive extension, geometry-based collision checking, mobility-based cost setting, and mobility-based analysis extension.

[0037] (3) An efficient and coordinated trajectory generation method is proposed. Path conflict coordination is a crucial final task in multi-UGV scenes. To balance efficiency, accuracy, and trajectory generation, this embodiment utilizes a novel tool: a physics engine, which allows for the free construction of a near-realistic world and the collection of near-realistic trajectories. Based on the collected trajectories, the 2D trajectories are discretized into a set of 3D spatiotemporal cube trajectories, thereby producing an efficient path coordination method.

[0038] Compared to other computer-aided design (CAD) software, BIM's main cartographic advantage lies in its semantic data. The most direct semantic information generated in map creation is the object type, which can be called an IFC entity. Furthermore, IFC's attribute definition mechanism allows users to customize attributes through attribute sets. For example, based on the customized attribute risk level of each BIM module, inaccessible spaces around existing obstacles can be updated to form a risk-oriented map. Similarly, occupied and unoccupied spaces (free space) on the map can be updated based on semantic status.

[0039] Given geometric and semantic information, a cell-based grid map is sufficient to decompose space into multiple free (unoccupied) or occupied cells. However, as another map type, network-based maps require additional topological information to represent the connectivity and reachability between subspaces. The module IfcRelSpaceBoundary is a generic IFC entity that determines the physical or virtual relationships between each space and its surrounding elements. By calculating the distance between the centroids of connected nodes in the network graph, geometrically meaningful conceptual boundaries can be updated. Topological information can also serve as a map simplification method to exclude irrelevant objects. For example, the module IfcRelContainedInSpatialStructure simplifies the entire map by saving only objects on specified floors.

[0040] The following examples establish UGV-integrated BIM for resource integration, and further propose a multi-layer map generation method to clearly construct and efficiently maintain maps.

[0041] After constructing the map, the Global Optimal Path (GPP) algorithm needs to be implemented to find the globally optimal path. Typical GPP algorithms can be divided into graph search-based GPP algorithms, sample-based GPP algorithms, and artificial intelligence-based GPP algorithms. The GPP algorithm used in this embodiment is an improved version of the traditional graph search-based GPP algorithm.

[0042] Although traditional graph-based search algorithms (such as the A* algorithm) can guarantee completeness, optimality, and efficiency, the A* algorithm still has four limitations because it does not combine the mobility and geometric characteristics of UGVs: (1) It only considers some UGV mobility systems (such as the Ackerman steering system) during expansion, and cannot be applied to all UGVs; (2) Secondly, the robot is generally converted into a single mesh for collision checking. Therefore, when the robot is non-circular, obstacles will be expanded by the robot's outer diameter, thereby reducing the configuration space (C-Space). Some studies have tried to solve this problem by expanding obstacles or calculating the actual footprint based on the robot's direction, but the problem of wasted mesh space still exists; (3) Thirdly, the mobility of UGVs is not well considered in the cost setting part; (4) Finally, the discretized control actions (mobility-based primitives) are difficult to end at the target position. In order to make up for the problem of search accuracy and speed up the search, a method called "Analytical Expansion (AE)" is proposed here. This method is based on the kinematic model and uses analytical expansion to expand the search. However, conventional methods require knowledge of the target location and orientation, while most on-site path planning tasks only provide the start and end points, leading to feasibility issues. Furthermore, previous Action Engineers (AEs) only concern themselves with the kinematics of specific robot states, lacking a complete UGV-centric AE. Therefore, the following embodiment designs an improved graph search-based Geometric Layout Planner (GPP) algorithm, incorporating UGV characteristics into various parts of the GPP algorithm to address the aforementioned limitations.

[0043] The basic hardware architecture of this embodiment will be introduced below. According to various embodiments of the present invention, Figure 1 A system 1 is provided, which includes a server (also known as a global path planning server) 100 and one or more unmanned ground vehicles (hereinafter referred to as UGVs) 200(1) to 200(N). The server 100 includes a processor 110, a non-volatile storage circuit 120 and a data communication circuit 130.

[0044] Non-transient storage circuit 120 is configured to store machine instructions (also called programs) 121. The program 121 includes the firmware of the server 100 and software / applications executing on the server 100. The non-transient storage circuit 120 may also store a database 122, which may be used to store data, including received data, established models, data generated by the processor 110, one or more digital map data, server information, user information, UGV information, and location information. The data communication circuit 130 is configured to receive data (e.g., input modeling data MD and designed task plan data TPD1) wirelessly or via wired means. Furthermore, the data communication circuit 130 is further configured to establish a wired and / or wireless network connection NC between the server 100 and the UGV 200 to transmit data (e.g., coordinated task plan data TPD2 and robot operating system (ROS) control data). The input modeling data MD includes a Building Information Modeling (BIM) model based on Industry Foundation Classes (IFC) and related parameters / information; the UGV model and related parameters / information for UGV 200; and the properties of UGV 200. The designed task planning data TPD1 includes default task planning data corresponding to different UGV 200s, indicating planned UGV movement information such as time, origin, destination, speed, etc.

[0045] Processor 110 executes machine instructions 121 to implement the provided method according to this embodiment.

[0046] Please refer to Figure 2 , Figure 3 and Figure 4 In one embodiment, the method implemented by server 100 includes steps S210 to S240.

[0047] Specifically, in step S210, processor 110 receives input modeling data and designed task planning data (B30) of the target unmanned ground vehicle (UGV). The input modeling data includes an IFC (Industry Foundation Course)-based BIM (Building Information Modeling) model B301 and an UGV model B302. Specifically, (1) the IFC-based BIM model (in .ifc format) is assumed to be available from AEC (Architecture, Engineering and Construction) managers and can be constructed using tools supported by IFC (such as Revit); (2) the source of the UGV 3D model is assumed to be provided by the robot supplier. These models are generally generated by robot modeling platforms such as SolidWorks or 3ds Max (which can be exported in .fbx format). The robot supplier also supports robot information, including geometry and dynamics (kinodynamics), usually in a fixed text or table format.

[0048] Next, in step S220, the processor 110 performs information integration (S311) and map generation (S312) based on the input modeling data using Building Information Modeling (BIM), thereby obtaining the UGV integrated BIM model B311 and the semantic map B312 of the interior space (e.g., Figure 3 The first part, B31, is shown.

[0049] Next, in step S230, the processor 110 sets the target UGV center path planning algorithm based on the UGV-integrated BIM model B311 and the semantic map B312, thereby generating the corresponding robot operating system (ROS) control data for the target UGV (e.g., Figure 3 The second part, B32, is shown.

[0050] In step S240, the processor 110 performs trajectory generation and task coordination based on the physics engine (PE) according to the UGV integrated BIM model, the semantic map, the set target UGV center path planning algorithm, and the designed task planning data, thereby obtaining the final UGV path of the target UGV. The final UGV path and the ROS control data are then sent to the target UGV (e.g., ...). Figure 3 The third part, B33, is shown.

[0051] In simple terms, the global path planning system provided by this invention is developed using resources such as the IFC-based BIM model B301, UGV model B302, and UGV attribute B303. More specifically, the main framework of the GPP system is as follows: Figure 4As shown, it mainly consists of three parts: (1) BIM-based information integration and map generation B31; (2) UGV center path planning B32; (3) Physical engine-based trajectory generation and coordination B33.

[0052] Please refer to the following instructions. Figure 4 Part 1: First, the IFC-based BIMB301, UGV model B302, and UGV attribute B303 are combined to create the IFC entity of the UGVs. The 3D model of the UGVs can be integrated into the IFC-based BIM platform (MK1) via an .ifc format file. Furthermore, UGV attribute B303 is collected and added to the existing BIM model via the "PropertySet" module of IFC-supported software. The constructed UGV-integrated BIMB311 serves as a knowledge base for map generation, and a hierarchical map generation method S312 is proposed to construct a semantically hierarchical map B312. The UGV-integrated BIM model B311 and semantic map B312 will later be imported into the physics engine PE (Part 3) for motion simulation. MK1 is marked with the IFC identifier.

[0053] Part Two: Before performing motion simulation, a path planning algorithm centered on UGVs (also referred to as the UGV-centered path planning algorithm) is proposed based on the traditional A* algorithm. The mobility of UGVs is analyzed and categorized into "unrestricted" and "restricted" types, and two mobility-based primitives are customized for each type (S331). Furthermore, to improve the accuracy and efficiency of collision checking, a geometry-based collision checking method is designed (S332). To ensure the energy efficiency of the path, the cost setting of the traditional UGV-centered path planning algorithm is improved (S333), further considering the mobility of UGVs (e.g., turning). The final step also considers mobility to improve the accuracy and search speed of path planning, referred to as "mobility-based analysis extension (S334)". Part Two will be incorporated into the UGV prototype in Part Three.

[0054] Part Three: Finally, a physics engine is selected as the platform for motion simulation and collision avoidance. The geometric portion of the UGV-integrated BIM model is imported into the physics engine (e.g., Unity3D) in .fbx format. Semantic portions (e.g., model attributes) are extracted from the IFC and saved in a fixed file format (e.g., Excel) via the IfcOpenShell module (Part One). Both geometric and semantic information can be read using the physics engine's built-in language (e.g., C# in Unity3D). Given the imported IFC resources, the required physical, kinematic, or dynamic modeling can be performed on the UGV prototype UP and environment prototype EP. The UGV prototype is further equipped with the constructed map (Part One) and the UGV-centric global path planning algorithm (Part Two), and the corresponding built-in language to perform motion simulation (S321), thereby collecting reliable trajectories (e.g., planned trajectory DTJ) during the simulation process. To avoid potential trajectory conflicts, a spatiotemporal coordination method (S322) is provided to discretize / decompose the 2D planar trajectory into multiple 3D continuously arranged space-time cubes, and to correct collisions according to user-defined priorities (e.g., the priority of each UGV) to obtain an adjusted trajectory ATJ. The MK2 designation represents Visual Studio Code (VSCode), a source code editor. Specifically, in this embodiment, C# is used to convert geometric and semantic information into code understandable by the physics engine to generate UGV and environment prototypes. For example, the "IfcOpenShell" module from the Python package is first used to extract the semantic parts and save them as an Excel file. Then, a programming language required by the physics engine (e.g., C#) is used to read the information stored in the Excel file and import it into the physics engine platform. The physics engine can then provide the model. The corresponding source code editor and code can be used to automate the provided methods, such as reading input, performing path planning and tracking, collecting trajectories, and coordinating collisions.

[0055] Please refer to Figure 5A Step S220 includes steps S221 to S223. Specifically, in step S221, the processor 110 integrates the Building Information Modeling (BIM) based on the industrial base classification with the UGV model of the target UGV, thereby constructing an integrated UGV BIM model. Next, in step S222, the processor 110 adds the UGV attributes of the target UGV to the integrated UGV BIM model by inputting modeling data.

[0056] Please refer to Figure 6 This embodiment extends BIM to BIM integrated with UGV, serving as a unified knowledge base for various UGV applications (such as map generation and prototype modeling). UGV entities are represented by the module IfcBuildingElementProxy, a generic IFC entity recommended for use with specialized machinery. The final IFC structure of the UGV entity is as follows... Figure 6 As shown, it can be built using IFC authoring tools (such as Revit, ArchiCAD).

[0057] The UGV model and its information are constructed accordingly as part of the IfcBuildingElementProxy (UGV) module. Most robotics software supports the .ifc format (such as SolidWorks), and models in .ifc format can be imported into software supported by IFC. Tessellated geometry is a common representation in robot modeling software, and this functionality is supported by the IFC module ifcshapereRepresentation (named "Tessellation"). Shapes are handled by the IfcPolygonalFaceSet module, where all faces are defined by polygons. Additionally, the IfcIndexedColourMap module is used to store color information for individual entities. The relational entity module IfcProductDefinitionShape is used to link the IfcShapeRepresentation module to UGV entities.

[0058] In addition to the UGV model, two UGV property sets need to be added: "Geometry" and "Kinodynamics." The geometry property set represents the robot's dimensions, while the kinematics property set includes kinematic and dynamic properties. These are defined by the IfcPropertySet module as data C_Pset_Geometry and C_Pset_Kinodynamics, and further linked to the UGV entity by the IfcRelDefinedByProperties module. Custom properties for each property set are represented by the IfcPropetySingleValue data. Typically, these IFC-related objects can be added to the UGV model to be imported using the PropertySet module of IFC-supported software (such as Revit).

[0059] Please return to Figure 5A In step S223, the processor 110, based on the UGV integrated BIM model (e.g., Figure 7A In B70, a layered map based on BIM is generated to obtain a semantic map related to the indoor space.

[0060] Please refer to Figure 7A In detail, a grid map is generated by tessell (decomposing) a planar space into a finite number of square cells to suit robot path planning. Unlike traditional techniques that use a single grid map, this embodiment proposes a multi-layer map generation method and improves upon it in two dimensions (e.g., Figure 7A (As shown). Horizontally, the relationships between IFC data and map components were established, improving data traceability. Vertically, the internal relationships and map sequences between different map components were defined. Specifically, the steps for generating a BIM-based layered map include: adding a lifecycle layer to the semantic map; adding a reality layer to the semantic map; and adding a UGV layer to the semantic map.

[0061] In more detail, the content of each layer and its corresponding IFC source are as follows:

[0062] First layer: Lifecycle layer. This layer contains all components of the current stage. Since each building project goes through several stages in its lifecycle, the IFC model may differ accordingly, including design-time BIM, construction-time BIM, Operation and Maintenance (O&M) BIM, demolition BIM, etc. This article takes O&M BIM as an example, consisting of walls, doors, columns, and slabs. First, the IfcBuildingStorey module is used for floor filtering to extract elements on the target floor. Then, elements from the IfcBuildingElement module (e.g., ...) are extracted. Figure 7A Extract the architectural elements of the selected floor from the sub-IFC entity (B71) of the layer, including elements IfcWall, IfcDoor, IfcColumn, and IfcSlab (e.g., the map of this layer is ML71). The size and position of each element can be inferred from the module ifcshapereRepresentation. Specifically, the data IfcBoundingBox is the minimum bounding box of each element; used to represent components with regular shapes, such as elements IfcColumn, IfcWall, and IfcDoor. The element IfcSweptAreaSolid is used to represent planar objects like element IfcSlab, which can be used to infer the coordinates of boundary points. Finally, the first layer can be determined to have, but is not limited to, the following characteristics:

[0063] (1) The map shape is a 2D bounding box containing the IfcSlab region, whose axis is parallel to the road surface (slab) axis.

[0064] (2) Set the unit grid size to the minimum width d of IfcWall. It is allowed to vary depending on the user's computing power (e.g., users with high computing power can use a smaller grid size).

[0065] (3) The mesh covered by IfcSlab is constructed as free space.

[0066] (4) Mesh that is completely or partially covered by elements other than IfcSlab is constructed as an obstacle.

[0067] (5) The global coordinates of each building element are saved as (X,Y), which can be converted into grid coordinates of (X / d,Y / d).

[0068] The second layer: the reality layer. This layer is responsible for updating reality changes, such as the status of doors and the congestion level of paths. This update information can be collected using BIM and wireless sensor networks. For example, suppose these reality updates have been measured by the installed sensors and stored in the BIM as module / element "IfcBuildingElementProxy(reality)" B72 (e.g., the map of this layer is ML72). Taking the door status as an example, the connected door entity is first located using the custom attribute "Door_ID(Pset)", and then "Door_Status(Pset)" is used to indicate whether the door is locked or unlocked.

[0069] The third layer: UGV layer. This final layer is designed to enable updates for UGVs, providing each UGV with a unique map. In the case of multiple UGVs, update efficiency can be improved when drawing this layer for each UGV, as the previous layers can be reused. First, the height of the UGV (the upper height of IfcSlab + C_Height) is compared with the bottom height of each obstacle (such as the bottom edge of a door), and the portion above the UGV is converted into free space. To ensure safe movement, free space inflations are added around the obstacles according to the UGV's module C_SafetyBuffer (e.g., the map for this layer is ML73). The relevant attributes of the UGVs used in the third layer (e.g., C_Height, C_SafetyBuffer) can be obtained from the UGV-integrated BIM model B70 via the module "IfcBuildingElementProxy(UGV)" B73.

[0070] Please refer to the following at the same time Figure 7AMap 7B illustrates the constructed maps MP71 for all-wheel-steering UGVs and MP72 for Ackerman-steering UGVs. The first layer selected the operation and maintenance phase, considering four main building elements: IfcSlab, IfcWall, IfcDoor, and IfcColumn. Grids covered by IfcSlab were marked as free / empty, while those covered by IfcWall, IfcDoor, and IfcColumn were marked as occupied. Furthermore, the spaces covered by IfcWall, IfcDoor, and IfcColumn were precisely constructed as polygonal barriers with explicit coordinates (for geometry-based collision checking). In the second layer, IfcBuildingElementProxy (Reality) data was used to update the door status, where all three doors were "unlocked" for both UGVs. Finally, on the map depicting the UGV layers, the height of the UGV was compared to the height of the open door to determine its passability. Additionally, safety buffer zones of 20 cm and 10 cm were added around the existing all-wheel-steering UGV and Ackerman-steering UGV barriers, respectively.

[0071] Please refer to Figure 5B Step S230 includes steps S231 to S234. Specifically, in step S231, the processor 110 performs a mobility-based primitive expansion to obtain one or more first location nodes (also called path points) of one or more UGV paths of the target UGV.

[0072] Next, in step S232, the processor 110 performs a geometry-based collision check to find one or more second location nodes in the first location node, wherein the second location nodes pass the geometry-based collision check.

[0073] Next, in step S233, the processor 110 sets a cost by taking into account the UGV mobility of the target UGV in order to find a target location node in the second location node, wherein the target location node has the minimum cost.

[0074] Next, in step S234, processor 110 executes a mobility-based analytical expansion (AE).

[0075] More specifically, the path planning algorithm module provided in this embodiment is a variant of the traditional A* algorithm, applicable to various UGV mobile systems, as shown in Algorithm 1 below.

[0076] The search loop is similar to the standard A* algorithm, where O and C represent the open set and closed set, respectively. The search process begins with the node with the minimum cost in the open set. During the primitive expansion (PrimitiveExpansion()), continuous motion primitives related to UGV movement are used as graph edges to generate new nodes, while nodes not in the closed set (C.contain()) are preserved. Then, the collision checking module (CollisionChecking()) checks the feasibility of the remaining nodes. The node with the minimum total cost, if not visited, is added to the open set via the cost setting modules (Primitivecost(), HeuristicCost()). To speed up the search and improve the final accuracy, the mobility-based analysis expansion module (UGV_AE()) is executed once when approaching the destination. The above process is repeated until any primitive reaches the destination.

[0077] Regarding mobility, please refer to Figure 8 Generally, UGV maneuvering systems can be categorized into (a) omnidirectional steering V81, (b) all-wheel steering V82, (c) Ackerman steering V83, (d) angular steering V84, and (e) sliding steering V85. (a) Omnidirectional UGVs supported by full wheels have the strongest capabilities, as their controllable degrees of freedom (DOF) are equal to their total degrees of freedom, allowing for instantaneous movement in any direction. (b) All-wheel steering systems consist entirely of rotating wheels; they can only move in any direction by turning the wheels to the required angle. Due to the decoupling of motion direction and body direction, both systems possess excellent ability to track paths of any shape. Therefore, this embodiment groups omnidirectional steering and all-wheel steering systems under "non-restrictive UGVs," directly employing primitives from the basic A* algorithm to fully explore the centers of eight neighborhoods. In addition to translation, non-restrictive UGVs also possess self-rotation capabilities, allowing entry into the same mesh in different postures. Therefore, by adding a set of discrete rotations (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°) during each primitive expansion process, the final primitive is as follows: Figure 9 As shown on page 91. Therefore, a set of 64 primitives can be expanded in each step. The pose update can be calculated by the following equations (1) to (4) (d is the unit mesh size), where (X,Y) and θ are the absolute directions of the world coordinates and the UGV pivot pose (e.g., Figure 8 (As shown). X′=X+ΔX,ΔX∈[-d,0,d] (1) Y′=Y+ΔY,ΔY∈[-d,0,d] (2) θ′=θ+Δθ,Δθ∈[0°,45°,90°,135°,180°,225°,270°,315°] (3) ΔX*ΔY≠0 (4)

[0078] Due to the presence of fixed wheels, the movement direction of three types of UGVs—(c) Ackerman Steering V83, (d) Corner Steering V84, and (e) Anti-Slip Steering V85—is restricted by the vehicle's orientation. Therefore, these three systems are classified as "restricted UGVs," and their kinematics can be simplified as follows: Figure 10 The bicycle model shown. The length of each primitive is set to d to ensure that only adjacent meshes are accessed, and the optimal solution is obtained in continuous space. It can be observed that during turning motion, feasible primitives are curves (e.g.,...). Figure 9 The new pivot posture can be calculated by equations (5) to (8), as shown by the black curve on page 92 of the diagram. m∈[-1,1](8)

[0079] Although the three types of restricted UGVs share the same kinematic model, they differ slightly in their turning radius constraints. Ackerman steering and angular steering both have a steering angle constraint α achieved through rotation. max To turn using the wheels (e.g.) Figure 8 (As shown in V83 and V84). To balance search efficiency and accuracy, the maximum turning distance is discretized into six parts to determine their radii r. On the other hand, the UGV with slip steering uses the speed difference between the wheels on both sides to turn, thus getting rid of the limitation on the turning radius. Similarly, to ensure the integrity and efficiency of the search, a uniform distribution set [±20°, ±40°, ±80°] is used. The details of the turning radius are described by the following equation (9).

[0080] Finally, the `.contain()` check module is used to determine whether to access the extended node. First, the world coordinates (X,Y) of the UGV axis center are converted to its corresponding mesh coordinates (x,y). Figure 11A The grid is then checked, and the grid position (x,y)((X / d,Y / d)) and logarithmic direction θ are examined to avoid duplicate nodes.

[0081] Please refer to Figure 11B In the traditional A* algorithm, the representation of the outer bounded circle may be too conservative and cannot pass through narrow corridors, especially for UGVs that have a long and narrow footprint. Furthermore, Figure 11AThe gray grid in the image is not completely covered by the UFV, which represents a waste of grid space. To address these issues, this embodiment designs a geometry-based collision detection method.

[0082] First, to ensure the accuracy of collision checking, a three-stage collision checking process was designed: Axis-Aligned Bounding Boxes Intersection (AABBI), Line-Line Intersection (LLI), and Point-in-Rectangle Intersection (PiRI). Here, a rectangular UGV profile is selected for illustration; other polygons can be checked using the same principle.

[0083] (1) AABBI check, used to check two axial boundary boxes (e.g., Figure 12A Whether the axial boundary boxes B121 and B122 in the diagram P121 intersect. An intersection will occur if their coners satisfy the following equations (10) and (11). A.minX≥B.maxX||A.maxX≤B.minX (10) A.minY≥B.maxY||A.maxY≤B.minY (11)

[0084] (2) LLI inspection aims to achieve better accuracy by testing whether any part of the UGV profile intersects with any part of the obstacle profile. For example, Figure 12A The two line segments (seg[(x1,y1),(x2,y2),seg[(x3,y3),(x4,y4)]) in the diagram P122 will intersect if they satisfy the following equations (12) to (14). (y4-y3)(x2-x1)-(x4-x3)(y2-y1)≠0 (12)

[0085] (3) PiRI check, used to check for a special collision situation that can evade LLI check, which occurs when one object is inside another object (e.g., Figure 12A(See illustration on page 123). First, the obstacle outline is divided into triangular segments. By using the following equations (15) to (17), if the corner of any UGV outline can be expressed as a positive linear combination of the corners of any triangular segment, then a collision can be determined. The situation where the obstacle is contained within the UGV should also be checked.

[0086] Please refer to Figure 12B In step S1200, processor 110 begins the collision check procedure. Next, in step S1210, processor 110 determines whether the AABBI check is passed. If yes, it determines that there is no intersection / collision (S1240); if no, it proceeds to step S1220. In step S1220, processor 110 determines whether the LLI check is passed. If yes, processor 110 further executes step S1230; if no, processor 110 determines that there is an intersection / collision (S1250). In step S1230, processor 110 determines whether the PiRI check is passed. If yes, it determines that there is no intersection / collision (S1240); if no, it determines that there is an intersection / collision (S1250).

[0087] However, performing a three-stage inspection on all obstacles could result in very high computational costs, especially in obstacle-filled indoor environments. Therefore, to improve detection efficiency, this embodiment proposes two acceleration methods:

[0088] (1) Integration with mesh-based inspection. While mesh-based inspection can be overly conservative and wasteful of space, it can still be used as a coarse-grained inspection. In other words, if all meshes covered by UGV ( Figure 11A If all the gray grids in the code pass the grid-based check (e.g., the grid covered by the UGV does not overlap with the obstacle's grid), then it is determined that no collision has occurred. If the grid-based check fails, the aforementioned precise three-stage check will be performed.

[0089] (2) Only check surrounding obstacles. Intuitively, collision checking does not need to check all obstacles in the space. That is, a check range can be set to reduce the amount of calculation. The size of the check range is recommended to be the sum of the length of the longest obstacle and the outer radius of the UGV.

[0090] In the terminology of the A* algorithm, g cThe actual cost is used to represent the optimal path from start to finish, and it consists of the cost of N primitives. In order to find a path that is optimal in both time and energy, the actual cost is designed to include the distance traveled, changes in direction, and changes in motion. Since the mobility of unrestricted UGVs and restricted UGVs is slightly different, their cost functions are different, as represented by the following equations (18) and (19).

[0091] For an unrestricted UGV, the movement distance d(n) indicates the length of the nth primitive, which can be calculated using equations (1) and (2). The change in direction (|θ(n) - θ(n-1)|) originates from self-rotation movement and is equal to Δθ in equation (3). Because the direction of movement and azimuth of movement made by an unrestricted UGV are decoupled, its steering cost s(n) is added to correspond to the change in direction of movement, and it can be determined based on the position of the next step (e.g., 45° for diagonal grids and 90° for transverse grids). Furthermore, Δθ and Δs are mutually exclusive because, intuitively, the existence of self-rotation movement eliminates the steering cost of the current node.

[0092] For a restricted UGV, d(n) is the element length calculated by equations (5) and (6). Since the body orientation and the movement direction of the restricted UGV are coupled, only the cost of orientation change is considered (equation (7)). In addition, switching the m(n)⊕m(n-1) orientation (e.g., from forward or reverse) incurs additional costs. The coefficients α and β are determined by the user.

[0093] The heuristic cost module HeuristicCost() in the A* algorithm can reduce the number of searches by providing promising directions to the destination. Unlike the primitive cost module PrimitiveCost(), which focuses on the movement characteristics of UGVs, the heuristic cost module HeuristicCost() focuses more on the environment (e.g., destination, obstacle avoidance). Therefore, without further UGV-related improvements, Euclidean distance is adopted as a general method for the heuristic cost module. Furthermore, since Euclidean-based heuristics are always acceptable, optimality can be guaranteed.

[0094] Please refer to Figure 13A and 13BTo compensate for the accuracy issues of motion-based primitives and accelerate path search, two analytical extensions are proposed for two UGV mobility systems:

[0095] (1) Non-restrictive UGV, which has excellent translational capabilities. Its Analytical Expansion (AE) is a straight line segment from the current position to the target position. The direction of the UGV will initially change towards the destination, and after the direction is confirmed, the direction will remain unchanged during the movement.

[0096] (2) For restricted UGVs, the analysis extension (AE) is based on the minimum turning radius. The nearest turning point c is calculated using the minimum turning radius. Two tangent points T1 and T2 can be found from the target G to the turning circle. Among the two candidate analysis extensions (U-T1-G, U-T2-G), the shorter AE is selected for collision checking. If the shorter AE fails (does not pass the collision check), the other is selected as an alternative, and a collision check is performed.

[0097] Finally, the discovered analysis is discretized into multiple pathpoints / location nodes for collision checking. To ensure accuracy, the resolution is set to be slightly smaller than the mesh cell size d. Furthermore, analysis extension is only allowed when the UGV is close to its destination to avoid a slow-down effect.

[0098] Please refer to Figure 5C Step S240 includes steps S241 to S245. In step S241, the processor 110 imports the geometric information data and semantic information data of the UGV-integrated BIM model into the physics engine.

[0099] Next, in step S242, the processor 110 models the UGV prototype and the indoor space environment prototype of the target UGV based on the imported geometric and semantic information data. The UGV prototype integrates a semantic map and a pre-set UGV center path planning algorithm. Next, in step S243, the processor 110 performs path planning and following operations based on the UGV prototype, the environment prototype, and the designed task planning data to obtain the first trajectory of the target UGV. Next, in step S244, the processor 110 performs spatiotemporal coordination operations based on the first trajectory of the target UGV and one or more further first trajectories of other UGVs to adjust the first trajectory into a second trajectory as the final UGV path of the target UGV. Next, in step S245, the processor 110 integrates the final UGV path into coordinated task planning data and sends the coordinated task planning data to the target UGV.

[0100] More specifically, the trajectory generation and coordination part based on the physics engine first models the prototype of the UGV and the prototype of the environment. The geometric models of the UGV and the built environment are exported from IFC-based BIM in .fbx format to the physics engine (e.g., Unity3D), and the format is converted through IFC software (e.g., Revit). The model attributes and the constructed map are saved in a fixed-format file (e.g., Excel) and are directly read by the physics engine.

[0101] The decoupling method for trajectory coordination requires less computation and is therefore more practical. Specifically, this embodiment chooses priority-based coordination. To effectively achieve spatiotemporal coordination, this embodiment proposes a novel space-time cube format to represent trajectories.

[0102] Please refer to Figure 5D Step S244 includes steps S2441 to S2444. In step S2441, the processor 110 generates multiple first spatiotemporal cubes corresponding to the target location nodes of the first trajectory based on the position coordinates and time of each target location node of the target UGV, thereby converting the first trajectory into a first 3D trajectory. Next, in step S2442, the processor 110 generates multiple further first spatiotemporal cubes corresponding to the further target location nodes of the further first trajectory based on the position coordinates and time of each further target location node of the other UGV, thereby converting the further first trajectory into a further first 3D trajectory. Next, in step S2443, the processor 110 adds the first spatiotemporal cubes and the further first spatiotemporal cubes to the 3D spatiotemporal coordinate space. Next, in step S2444, in response to determining that the target first spatiotemporal cube in the first spatiotemporal cube in the 3D spatiotemporal coordinate space overlaps with one of the further first spatiotemporal cubes, the processor 110 adjusts the time and / or coordinates of the target first spatiotemporal cube to update the target first spatiotemporal cube as the target second spatiotemporal cube, thereby adjusting the first trajectory to the second trajectory as the final UGV path, wherein the target second spatiotemporal cube corresponding to the target UGV does not overlap with any of the further first spatiotemporal cubes corresponding to other UGVs.

[0103] More specifically, the details of the proposed spatiotemporal trajectory coordination algorithm are shown in Algorithm 2 below.

[0104] After modeling / simulation in the physics engine, trajectory information can be collected (line 2 of the algorithm). Given the size of the UGV and the time-stamped poses of the UGV pivot point, the corresponding occupied mesh can be inferred. Figure 14A The time slot occupied by each grid can also be recorded (as shown in the OT in program (21) below). U1i This allows us to extend 2D geometric paths to 3D spatiotemporal cube trajectories. Figure 14B The uncoordinated trajectory set U (Equation (20)) is sorted based on priority (line 4 of the algorithm), and the priority can be determined by the administrator. Before coordination, the first priority path t is moved from U to the coordinated trajectory set C (lines 5 and 6 of the algorithm), which does not need to be adjusted. U = [U1, U2, ..., U n (20)

[0105] For sorted paths, identifying spatial and temporal conflicts is crucial. If U is not empty, the current first-priority trajectory u is extracted for use with each coordinated trajectory c. i Perform conflict analysis (lines 8 and 9 of the algorithm). For example, Figure 15A As shown, the overlapping region CR in the 2D coordinate space represents all potential regions that may cause conflict. Two paths share a common grid [X,Y] (line 10 of the algorithm). For the 3D spatiotemporal coordinate space, time slots of the same 2D conflicting grid need to be integrated into the occupied time slot set OT (line 11). Figure 15B as well as Figure 16 As shown in the first part, OT is the union of the occupied slots of all coordinated trajectories (e.g., OT[x,y] is all the occupied slots of grid [x,y]). Figure 15B as well as Figure 16 As shown in the second part, the empty timeslot set (ET) of each grid can be inferred as the complementary set of OT (line 15 of the algorithm). Each part of ET represents a feasible time slot for time adjustment. Furthermore, because UGVs are allowed to enter the same grid in different poses, each inspected grid may contain multiple occupied time slots. Figure 16 (Part 3). Therefore, the adjustable time range AT for each grid should be determined by comparing the occupied time slot ut. i Length and space slot et jThe length is determined (algorithm lines 14-17). The final adjustable time set FT is obtained from the intersection of the adjustable time ranges AT of all grids (line 19), and the final adjustment value is the smallest absolute value in the final adjustable time set FT (lines 23-28). Finally, the currently checked trajectory u is inserted into the coordinated set C (line 29), and the entire process is repeated until u is empty (all trajectories are coordinated).

[0106] Based on the above embodiments, this invention proposes a Global Path Planning (GPP) system for unmanned indoor ground vehicles (UGVs) based on Building Information Modeling (BIM) and a physics engine. First, by customizing the IFC structure, UGV configuration information is integrated into the BIM as a knowledge base for the GPP system, and a multi-layer map generation method with improved logic and efficiency is proposed. Second, considering the characteristics of UGVs, a UGV-centric pathplanning algorithm is designed, including mobility-based primitive expansion, geometry-based collision checking, mobility-based cost setting, and mobility-based analysis expansion. Finally, a reliable trajectory generation method is designed based on the physics engine, and an efficient spatiotemporal coordination method for collision avoidance is proposed. Compared with traditional path planning algorithms, this algorithm can reduce trajectory time and steering jitter by nearly 50% while maintaining computational speed, thereby improving the travel efficiency of the planned path.

[0107] Furthermore, this embodiment includes the following features:

[0108] By integrating UGV into BIM resources, interoperability can be achieved. In actual construction projects, robot-related information is generally stored in specialized manuals, which are often complex and unnecessary for laypeople, especially AEC users who are more concerned with GPP tasks and have little experience with robots. Furthermore, information retrieval can be very time-consuming when using different robot software and formats within the same project. With the help of the IFC model, a general IFC structure for UGV was designed to store the necessary information (geometric and semantic information). Integrating UGV into BIM will significantly improve the interoperability of UGV resources in actual projects, as it avoids cumbersome setup and redundant resources in robot applications. The integration principle of UGV-BIM can also be extended to the integration of other robots (such as drones and robotic dogs) and other types of information (such as the relationship between UGV and building elements).

[0109] BIM-based map generation improves cartographic efficiency. Typically, UGVs build their own navigation maps in real-world environments using SLAM. However, SLAM-based maps may be incomplete in terms of geometric (sensor noise) and semantic (single data attribute) information. With the increasing digitalization of the AEC (Advanced Environment Configuration and Control) industry, most indoor environments are now equipped with accurate BIM prototypes. Therefore, the proposed BIM-based UGV map generation method is expected to improve the efficiency of UGV map generation. Based on its precise positioning, the real UGV can directly read the BIM model lifecycle layer in a specific format (e.g., .yaml). The data source for the reality layer can be collected from IoT applications to update the site environment in real time. Furthermore, the UGV layer can handle multiple types of UGVs, and due to the shareable map layer in this embodiment, significant computational resources can be saved.

[0110] The algorithm enhances the global path planning capabilities of UGVs in real-world sites. In actual projects, the algorithm can be deployed to UGVs by writing a global planner plugin in the Robot Operating System (ROS) navigation stack. Furthermore, experimental results on the physics engine demonstrate that the UGV center A* algorithm exhibits high accuracy and computational speed. Therefore, it is expected that the deployment of the UGV center A* algorithm will significantly improve the global path planning capabilities of UGVs in real-world indoor sites, such as navigating narrow spaces and saving travel distance.

[0111] Physics-based engine coordination optimizes UGV mission plans before operation. Beyond serving as a validation platform (e.g., validating map generation and path planning algorithms), the physics engine here acts as a reliable trajectory generation method. With the increasing automation of the autonomous vehicle industry, more and more driverless vehicles (UGVs) are expected to appear in areas with a high probability of road collisions. Therefore, AEC (Autonomous Vehicle Engine) managers strongly encourage physics-based trajectory coordination before daily operations. After inputting the robot's mission plan into the physics engine, a pre-designed UGV prototype can perform motion simulations and output a set of reliable trajectories. Thus, potential conflicts can be identified and resolved in advance before the UGV is actually put into operation.

[0112] The functional units of the devices and methods according to the embodiments disclosed herein may be implemented using computing devices, computer processors, or electronic circuit systems, including but not limited to application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and other programmable logic devices configured or programmed according to the teachings of this disclosure. Machine instructions or firmware / software code running in the computing device, computer processor, or programmable logic device can be readily created by those skilled in the art based on the teachings of this disclosure.

[0113] All or part of the methods according to the embodiments can be executed in one or more computing devices including server computers, personal computers, laptop computers, mobile computing devices (e.g., smartphones) and tablet computers.

[0114] The embodiments include non-transitory memory circuitry and / or computer storage media having machine instructions or firmware / software code stored therein, which can be used to program a processor to perform any of the processes of the present invention. The non-transitory memory circuitry and / or storage media include (but are not limited to) floppy disks, optical disks, Blu-ray discs, DVDs, CD-ROMs and magneto-optical disks, ROMs, RAMs, flash memory devices, or any type of medium or device suitable for storing instructions, code, and / or data.

[0115] Each of the functional units according to various embodiments may also be implemented in a distributed computing environment and / or cloud computing environment, wherein one or more processing devices interconnected via communication networks such as intranets, wide area networks (WANs), local area networks (LANs), the Internet, and other forms of data transmission media execute all or part of machine instructions in a distributed manner. The communication network established by the data communication circuitry in various embodiments supports various communication protocols, such as (but not limited to) Wi-Fi, Global System for Mobile Communications (GSM), Personal Handheld Telephone System (PHS), Code Division Multiple Access (CDMA), Global Microwave Access Interoperability (WiMAX), third-generation wireless communication technology (3G), fourth-generation wireless communication technology (4G), fifth-generation wireless communication technology (5G), Long Term Evolution (LTE), Bluetooth, and Ultra Wideband (UWB).

[0116] The foregoing description of the invention has been provided for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations will be apparent to those skilled in the art.

[0117] The embodiments were selected and described in order to best illustrate the principles of the invention and its practical application, thereby enabling others skilled in the art to understand the various embodiments of the invention and the various modifications suitable for particular uses upon careful consideration.

Claims

1. A computer-based method for performing global path planning for unmanned ground vehicles (UGVs) in indoor spaces using a global path planning server, leveraging Building Information Modeling (BIM) and a physical engine, characterized in that... include: Receive input modeling data and designed task planning data for the target UGV; Based on the input modeling data, information integration and map generation are performed using Building Information Modeling (BIM), resulting in a UGV-integrated BIM model and a semantic map of the interior space. Based on the BIM model integrated with the UGV and the semantic map, a target UGV center path planning algorithm is set, wherein setting the target UGV center path planning algorithm includes: Perform mobility-based primitive extensions to obtain one or more first location nodes of one or more UGV paths of the target UGV; Perform a geometry-based collision check to find one or more second location nodes in the first location node, wherein the second location nodes pass the geometry-based collision check; Cost settings are made by considering the UGV mobility of the target UGV to find a target location node among the second location nodes, wherein the target location node has the minimum cost; and Perform mobility-based analytics extensions; By writing a global planner plugin in the robot operating system navigation stack, the target UGV center path planning algorithm is deployed to the target UGV, so that the global planner plugin can generate robot operating system control data corresponding to the target UGV; and Based on the BIM model integrated with the UGV, the semantic map, the set target UGV center path planning algorithm, and the designed task planning data, a physics engine-based trajectory generation and task coordination are performed to obtain the final UGV path of the target UGV. The final UGV path and the robot operating system control data are then sent to the target UGV.

2. The method according to claim 1, characterized in that, The steps for performing the information integration based on Building Information Modeling (BIM) and the map generation include: The Building Information Modeling (BIM) based on industrial base classification is integrated with the UGV model of the target UGV to construct an integrated UGV BIM model. Using the input modeling data, the UGV attributes of the target UGV are added to the UGV integrated BIM model; and Based on the BIM model integrated by UGV, a layered map based on BIM is generated to obtain a semantic map related to the indoor space.

3. The method according to claim 2, characterized in that, The steps involved in generating a BIM-based layered map include: Add a lifecycle layer to the semantic map; Add a reality layer to the semantic map; and Add the UGV layer to the semantic map.

4. The method according to claim 1, characterized in that, The steps for trajectory generation and task coordination based on the physics engine include: Import the geometric and semantic information data of the UGV-integrated BIM model into the physics engine; Based on the imported geometric and semantic information data, a prototype of the target UGV and an environmental prototype of the indoor space are modeled, wherein the UGV prototype integrates the semantic map and the set target UGV center path planning algorithm; Based on the UGV prototype, the environment prototype, and the designed task planning data, path planning and following operations are performed to obtain the first trajectory of the target UGV; A spatiotemporal coordination operation is performed based on the first trajectory of the target UGV and one or more other first trajectories of other UGVs, thereby adjusting the first trajectory into a second trajectory as the final UGV path of the target UGV; and The final UGV path is integrated into coordinated task planning data, and the coordinated task planning data is sent to the target UGV.

5. The method according to claim 4, characterized in that, The steps for performing the spatiotemporal coordination operation include: Based on the position coordinates and time of each target location node of the target UGV, multiple first spatiotemporal cubes corresponding to the target location nodes of the first trajectory are generated, thereby converting the first trajectory into a first 3D trajectory. Based on the position coordinates and time of each further target location node of other UGVs, multiple further first spatiotemporal cubes corresponding to the further target location nodes of the further first trajectory are generated, thereby converting the further first trajectory into a further first 3D trajectory; Add the first spacetime cube and the further first spacetime cube to the 3D spacetime coordinate space; and In response to determining that the target first spatiotemporal cube in the first spatiotemporal cube in the 3D spatiotemporal coordinate space overlaps with one of the further first spatiotemporal cubes, the time and / or coordinates of the target first spatiotemporal cube are adjusted to update the target first spatiotemporal cube as the target second spatiotemporal cube, thereby adjusting the first trajectory to the second trajectory as the final UGV path, wherein the target second spatiotemporal cube corresponding to the target UGV does not overlap with any of the further first spatiotemporal cubes corresponding to other UGVs.

6. A server that utilizes Building Information Modeling (BIM) and a physics engine to perform global path planning for unmanned ground vehicles (UGVs) in indoor spaces, characterized in that... include: Processor, wherein the processor is configured to execute a plurality of machine instructions to implement a global path planning method for UGVs in indoor spaces, the method comprising: Receive input modeling data and pre-designed task planning data for the target unmanned ground vehicle (UGV); Based on the input modeling data, information integration and map generation are performed using Building Information Modeling (BIM), resulting in a UGV-integrated BIM model and a semantic map of the interior space. Based on the BIM model integrated with the UGV and the semantic map, a target UGV center path planning algorithm is set, wherein setting the target UGV center path planning algorithm includes: Perform mobility-based primitive extensions to obtain one or more first location nodes of one or more UGV paths of the target UGV; Perform a geometry-based collision check to find one or more second location nodes in the first location node, wherein the second location nodes pass the geometry-based collision check; Cost settings are made by considering the UGV mobility of the target UGV to find a target location node among the second location nodes, wherein the target location node has the minimum cost; and Perform mobility-based analytics extensions; By writing a global planner plugin in the robot operating system navigation stack, the target UGV center path planning algorithm is deployed to the target UGV, so that the robot operating system control data corresponding to the target UGV can be generated through the global planner plugin; and Based on the BIM model integrated with the UGV, the semantic map, the set target UGV center path planning algorithm, and the designed task planning data, a physics engine-based trajectory generation and task coordination are performed to obtain the final UGV path of the target UGV. The final UGV path and the robot operating system control data are then sent to the target UGV.

7. The server according to claim 6, characterized in that, The steps for performing the information integration based on Building Information Modeling (BIM) and the map generation include: The Building Information Modeling (BIM) based on industrial base classification is integrated with the UGV model of the target UGV to construct an integrated UGV BIM model. Using the input modeling data, the UGV attributes of the target UGV are added to the UGV integrated BIM model; and Based on the BIM model integrated by UGV, a layered map based on BIM is generated to obtain a semantic map related to the indoor space.

8. The server according to claim 7, characterized in that, The steps involved in generating a BIM-based layered map include: Add a lifecycle layer to the semantic map; Add a reality layer to the semantic map; and Add the UGV layer to the semantic map.

9. The server according to claim 6, characterized in that, The steps for trajectory generation and task coordination based on the physics engine include: Import the geometric and semantic information data of the UGV-integrated BIM model into the physics engine; Based on the imported geometric and semantic information data, a prototype of the target UGV and an environmental prototype of the indoor space are modeled, wherein the UGV prototype integrates the semantic map and the set target UGV center path planning algorithm; Based on the UGV prototype, the environment prototype, and the designed task planning data, path planning and following operations are performed to obtain the first trajectory of the target UGV; A spatiotemporal coordination operation is performed based on the first trajectory of the target UGV and one or more other first trajectories of other UGVs, thereby adjusting the first trajectory into a second trajectory as the final UGV path of the target UGV; and The final UGV path is integrated into coordinated task planning data, and the coordinated task planning data is sent to the target UGV.

10. The server according to claim 9, characterized in that, The steps for performing the spatiotemporal coordination operation include: Based on the position coordinates and time of each target location node of the target UGV, multiple first spatiotemporal cubes corresponding to the target location nodes of the first trajectory are generated, thereby converting the first trajectory into a first 3D trajectory. Based on the position coordinates and time of each further target location node of other UGVs, multiple further first spatiotemporal cubes corresponding to the further target location nodes of the further first trajectory are generated, thereby converting the further first trajectory into a further first 3D trajectory; Add the first spacetime cube and the further first spacetime cube to the 3D spacetime coordinate space; and In response to determining that the target first spatiotemporal cube in the first spatiotemporal cube in the 3D spatiotemporal coordinate space overlaps with one of the further first spatiotemporal cubes, the time and / or coordinates of the target first spatiotemporal cube are adjusted to update the target first spatiotemporal cube as the target second spatiotemporal cube, thereby adjusting the first trajectory to the second trajectory as the final UGV path, wherein the target second spatiotemporal cube corresponding to the target UGV does not overlap with any of the further first spatiotemporal cubes corresponding to other UGVs.

11. A system for performing global path planning for unmanned ground vehicles (UGVs) in indoor spaces using Building Information Modeling (BIM) and a physical engine, characterized in that, include: Server; and Target UGV, The server's processor is configured to execute multiple machine instructions to implement a global path planning method for the target UGV in the indoor space, the method comprising: Receive the input modeling data and designed task planning data of the target UGV; Based on the input modeling data, information integration and map generation are performed using Building Information Modeling (BIM), resulting in a UGV-integrated BIM model and a semantic map of the interior space. Based on the BIM model integrated with the UGV and the semantic map, a target UGV center path planning algorithm is set, wherein setting the target UGV center path planning algorithm includes: Perform mobility-based primitive extensions to obtain one or more first location nodes of one or more UGV paths of the target UGV; Perform a geometry-based collision check to find one or more second location nodes in the first location node, wherein the second location nodes pass the geometry-based collision check; Cost settings are made by considering the UGV mobility of the target UGV to find a target location node among the second location nodes, wherein the target location node has the minimum cost; and Perform mobility-based analytics extensions; By writing a global planner plugin in the robot operating system navigation stack, the target UGV center path planning algorithm is deployed to the target UGV, so that the robot operating system control data corresponding to the target UGV can be generated through the global planner plugin; and Based on the BIM model integrated with the UGV, the semantic map, the set target UGV center path planning algorithm, and the designed task planning data, a physics engine-based trajectory generation and task coordination are performed to obtain the final UGV path of the target UGV. The final UGV path and the robot operating system control data are then sent to the target UGV.

Images