Robot semantic map management and navigation system

Abstract

The invention discloses a robot semantic map management and navigation system, and aims to solve the problems of non-visual navigation, low map management efficiency, inflexible system, poor expansibility and the like caused by insufficient semantic information utilization and management of an existing robot. The system comprises a unified map storage unit for acquiring semantic information in a mode of combining real-time identification and user definition, and a semantic navigation processing unit for performing navigation based on the semantic information; an automated map lifecycle management unit; a virtual obstacle management unit; and a plurality of function processing modules which adopt a modular architecture and realize communication through a self-defined ROS2 message type interface, and support plug-in navigation algorithm extension. According to the method, visual semantic navigation is realized, and the user experience is improved; the use of map resources is optimized, and the operation efficiency of the system is improved; and the navigation flexibility, the system expandability, the environment adaptability and the maintainability of semantic data are obviously enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of mobile robots, and more specifically, to a robot map management and navigation system based on semantic information. Background Art

[0002] With the rapid development of artificial intelligence and robot technology, the application of mobile robots has become increasingly popular in many fields such as industrial production, medical services, home assistance, and security patrol. In the prior art, robot systems usually use two-dimensional or three-dimensional geometric maps for environmental modeling and navigation. Such geometric maps, such as grid maps, mainly record the physical structure information of the environment, such as walls, fixed obstacles, etc. The robot senses the surrounding environment through sensors (such as lidar, camera) carried on it, and uses technologies such as SLAM (Simultaneous Localization and Mapping) to construct such geometric maps, and then performs path planning and positioning navigation based on the map.

[0003] However, the prior art solutions have many deficiencies: Insufficient utilization of semantic information leads to unintuitive navigation interaction: The existing systems mainly rely on precise geometric coordinates for navigation, resulting in poor user experience and unable to meet the intuitive navigation requirements based on natural language or object / area names (such as "go to the kitchen", "find meeting room A"). The system lacks in-depth understanding and utilization of high-level semantic information such as "objects", "rooms" in the environment. Low map management efficiency and high resource occupancy: As the robot continues to run, the map data increases continuously, resulting in increased storage and processing pressure. The prior art generally lacks an automated map lifecycle management mechanism, and a large amount of historical map data accumulates, which not only occupies valuable storage resources, increases the management complexity, but also makes it difficult to ensure that the robot always uses the latest and most effective map version. Rigid system architecture and poor scalability: The architecture design of traditional robot systems has a low degree of modularization, and the coupling degree between functional modules is relatively high, which is not conducive to system maintenance, function expansion, and integration with cloud computing platforms. Insufficient flexibility of navigation constraints: It is difficult for the prior art to flexibly set virtual obstacles (such as virtual walls, restricted areas) according to actual needs, which limits the adaptability of robot navigation and the response ability to temporary environmental changes. Limited semantic information processing and management capabilities: Even if some prior art begins to attempt to introduce semantic information, it mostly focuses on the visual recognition level and lacks in-depth integration with map management and navigation systems. In addition, the semantic categories are usually pre-fixed and cannot adapt to specific environments or special objects or areas defined by users. More importantly, the prior art generally lacks means for effectively managing (such as adding, deleting, modifying, querying) the annotated semantic information in the map.

[0004] In view of the above technical problems, we propose a robot system that can deeply integrate and manage semantic information, provide intuitive semantic navigation, and has the capabilities of automated map management and flexible navigation constraints. Summary of the Invention

[0005] The present invention aims to overcome the above-mentioned defects of the prior art and provides a robot semantic map management and navigation system.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A robot semantic map management and navigation system, the system includes:

[0008] Unified map storage unit: This unit is configured to store the map data sensed by the robot through sensors in the working environment. These map data include not only the geometric map information (such as raster data) describing the physical structure of the environment, but also the semantic information associated with this geometric map information. The semantic information is the semantic description and attributes of objects (such as tables, chairs, charging piles) or areas (such as kitchens, meeting rooms) in the environment. Part of these semantic information comes from the semantic tags automatically generated by real-time semantic recognition of the data of the robot sensors (such as RGBD cameras), and the other part comes from the semantic tags custom-added or edited by the user through the user interface (such as the WEB-side personalized semantic definition interface).

[0009] Semantic mapping module: This module is configured to execute the simultaneous localization and mapping (SLAM) algorithm based on the sensing data of the robot sensors (especially the RGBD camera that needs to obtain spatial coordinates) to generate the aforementioned geometric map information. Further, this module integrates a semantic recognition function, which can be accelerated by using, for example, a graphics processing unit (GPU). For the semantic objects that already exist in the preset model library of the system, these objects can be recognized in real time and automatically during the mapping or environment perception process, and their corresponding semantic tags and spatial position information can be generated or updated in the map.

[0010] Semantic navigation processing unit: This unit is connected to the unified map storage unit. When receiving a navigation instruction based on semantic information issued by the user or the upper-layer application (such as "go to the kitchen" or "go to the side of the charging pile"), this unit can parse the semantic target name in the instruction, query the semantic information stored in the unified map storage unit to find the specific position (such as the center point of the object or the preset navigation recommended point) and attitude of the corresponding semantic target, and based on the determined navigation target position and attitude, call the underlying path planning and motion control interfaces to control the robot to execute the navigation task.

[0011] Automated Map Lifecycle Management Unit: This unit is also connected to the Unified Map Storage Unit and is responsible for the automated management of the stored map resources. It monitors the number of maps stored in the Unified Map Storage Unit. When the addition of new maps causes the total number of maps to exceed a preset threshold, this unit will automatically identify and process (such as deleting or archiving) historical map data according to preset policies (for example, based on the timestamp of map creation, giving priority to deleting the oldest maps; or based on usage frequency, deleting the least frequently used maps), and can selectively back up the latest or most important maps, thereby optimizing storage resources and ensuring that the system uses valid maps.

[0012] Virtual Obstacle Management Unit: This unit allows users to define virtual obstacle information on the constructed map through a user interface (such as a WEB-based map editing interface), such as virtual walls (usually represented as line segments) or virtual restricted areas (usually represented as closed polygons). These user-defined virtual obstacle information will be updated and stored in the Unified Map Storage Unit as part of the semantic information. When the semantic navigation processing unit performs navigation path planning, these virtual obstacle information will serve as additional navigation constraints, and the path planner will identify and avoid the areas defined by these virtual obstacles.

[0013] System Scheduling Unit and Multiple Independent Functional Processing Modules: The system adopts a modular architecture design. In addition to the partial functions of the above core units / modules that can be implemented by these modules, it also includes a core system scheduling unit and multiple relatively independent and pluggable functional processing modules. These functional processing modules at least include a map operation module (responsible for map loading, switching, querying, virtual obstacle setting, etc.), a mapping operation module (responsible for starting / stopping mapping, saving maps, etc.), and a navigation control module (responsible for target point navigation, semantic target navigation, navigation control, etc.) that is part of the semantic navigation processing unit. The system scheduling unit is designed using, for example, the strategy pattern and can execute corresponding strategies according to different operation instructions issued externally (such as from the cloud or other high-level management systems) or internally, and then coordinate and schedule the relevant functional processing modules. The communication between these functional processing modules and between them and the system scheduling unit is achieved through a unified interface protocol based on custom ROS2 (Robot Operating System 2) message types. For example, specific message files (.msg) are defined for map operations, mapping operations, and navigation operations respectively, and these message files are included in a custom service file (.srv) to enable stable and efficient communication with cloud services or other external systems.

[0014] Preferably, during the semantic recognition process, the semantic mapping module uses a Graphics Processing Unit (GPU) for acceleration processing.

[0015] Preferably, the unified map storage unit specifically uses a database to store map data, and the semantic information is seriallyized in JSON format and stored in association with the geometric map data.

[0016] Preferably, the semantic information further includes object semantic information, which records the name, position, size, direction, navigation recommendation points, and category of objects in the environment; the navigation recommendation points are automatically calculated and generated at the completion stage of semantic mapping, and are located at a preset or user-modifiable specified distance in the direction of the object.

[0017] Preferably, when the automated map lifecycle management unit deletes a map that meets a specific policy, the specific policy is to delete the earliest created map based on the timestamp and automatically back up the latest map.

[0018] Preferably, the virtual obstacles defined by the virtual obstacle management module include virtual walls, which are represented as line segments, and / or virtual restricted areas, which are represented as polygons.

[0019] Preferably, the unified interface protocol based on the custom ROS2 message type includes specific message files (.msg) defined for map operations, mapping operations, and navigation operations respectively, and these message files are included in a custom service file (.srv) for communication with cloud services.

[0020] Preferably, it further includes a semantic data management interface unit that allows users to perform addition, deletion, modification, and query operations on the semantic information stored in the unified map storage unit through the user interface; when the user customizes a new semantic category, if the defined name duplicates an existing category name, the system will prompt and prevent saving.

[0021] Preferably, the navigation control module supports plug-in navigation algorithm extension, and different navigation algorithms can be selected and configured through a configuration parameter file.

[0022] Preferably, it further includes a cloud service interface for communicating with cloud services, and the cloud services can be connected to the WEB end and the personalized semantic definition interface.

[0023] Advantageous Effects Compared with the prior art, the present invention has the following advantageous effects:

[0024] 1. Deep semantic integration: Different from only recognizing semantics at the visual level or using semantics as simple annotations on the map, this solution deeply integrates semantic information (objects, rooms, virtual obstacles) with the geometric map at the storage layer (unified storage in the database), management layer (processing through a unified interface), and application layer (semantic navigation, semantic constraints). Users can use natural language and a visual interaction interface to replace traditional coordinate point navigation, significantly reducing the usage threshold and enhancing the user experience.

[0025] 2. Optimized system resource utilization: Through automated map lifecycle management, the system can intelligently control the number of maps, avoid excessive occupation of storage space, and ensure the long-term stable operation of the system.

[0026] 3. Unified modular architecture: Adopting a clear process / modular management architecture, decoupling the map, mapping, and navigation functions, improving the maintainability, scalability, and robustness of the system.

[0027] 4. Enhanced flexibility of the navigation system: The virtual obstacle setting function makes the restriction of the robot's activity area more flexible, and the navigation constraints can be dynamically adjusted without modifying the physical environment.

[0028] 5. Application-oriented semantic navigation: It not only realizes the acquisition of semantic information, but more importantly, provides the navigation execution ability based on semantic names, directly facing user needs.

[0029] 6. Improved environmental adaptability of the system: Users can customize semantic categories (such as "charging pile", "disinfection station") according to specific application scenarios, making the map more in line with actual needs.

[0030] 7. Enhanced maintainability of semantic data: Provide convenient interfaces to add, delete, modify, and query semantic information in the map, facilitating users to manage and update semantic information in the map, and maintaining the accuracy and timeliness of the data.

[0031] 8. Support for remote advanced tasks: Through the cloud integration interface, complex task instructions from a remote control center or an advanced scheduling system can be received, expanding the application scenarios of the robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the system architecture of the robot semantic map management and navigation system of the present invention.

[0033] Figure 2 It is a schematic diagram of the semantic map data and storage structure of the robot semantic map management and navigation system of the present invention.

[0034] Figure 3 It is a schematic diagram of the semantic navigation process of the robot semantic map management and navigation system of the present invention.

[0035] Figure 4 This is a schematic diagram of the map life cycle management process of the robot semantic map management and navigation system of the present invention.

[0036] Figure 5 This is a schematic diagram of the system component relationship of the robot semantic map management and navigation system of the present invention, including the semantic mapping module and map service interaction.

[0037] Figure 6 This is a schematic diagram of the detailed interaction data signal flow between modules of the robot semantic map management and navigation system of the present invention. Detailed implementation manners

[0038] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0039] To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and the accompanying drawings.

[0040] Collaborative work of the overall system architecture and core modules: Refer to Figure 1 , Figure 5 and Figure 6 , a robot semantic map management and navigation system provided by an embodiment of the present invention. This system is preferably built on the framework of the robot operating system (such as ROS2) to utilize the communication, device driver, and tool ecosystem provided by it. As Figure 1 shown, the system may include a core system scheduling unit, and a series of functional processing modules managed and coordinated by this scheduling unit. These functional processing modules mainly include:

[0041] The map editing processing module (i.e., the map operation module), the mapping instruction processing module (i.e., the mapping operation module), and the navigation processing module (i.e., the navigation operation module). The system scheduling unit is constructed using the strategy pattern in object-oriented design. This design pattern enables the system scheduling unit to flexibly select and execute corresponding processing strategies according to different types of operation instructions received (e.g., remote instructions from the cloud or requests from the local user interface). Each strategy encapsulates the logic for interacting with one or more specific function processing modules internally. These function processing modules run as independent processes, and their communication with each other and with the system scheduling unit is achieved based on the ROS2 standard communication mechanisms (such as Topics, Services, Action sequences) combined with custom ROS2 message types. Specifically, specific message files (e.g., MapOption.msg, MappingOption.msg, NavigationOption.msg) can be designed for core functions such as map operation, mapping operation, and navigation operation respectively, and these message files are encapsulated in custom service files (.srv). This approach not only standardizes the communication protocol between modules but also enables the system to conveniently exchange data and interact with external systems, enhancing the openness and integration capabilities of the system. Refer to Figure 6 , external instructions (such as those from the cloud service interface or the local user interface) first reach the system core scheduling node. The scheduling node parses the instructions and selects the corresponding strategy for processing according to the internal strategy pattern logic. This processing strategy will then interact with the corresponding function processing module (such as the map editing processing module, the mapping instruction processing module, or the navigation processing module) through the defined unified interface protocol (e.g., calling a specific ROS2 Service encapsulating custom ROS2 message types, or publishing a Topic containing custom ROS2 messages, or starting a ROS2 Action). For example, an instruction of "start mapping" may trigger an Action call to the mapping instruction processing module through the scheduling node, and this module will publish its status through a Topic and communicate with the map storage interface through a Service to save the map during execution. Similarly, a semantic navigation request will be routed to the navigation processing module through the scheduling node, and this module will query the unified map storage unit and control the movement of the robot.

[0042] Construction, Storage, and Management of Semantic Maps: Acquisition of Semantic Information and Semantic Mapping Module: In this system, the acquisition of semantic information adopts a strategy that combines real-time automatic recognition and user assistance to balance automation efficiency and information completeness. The core semantic mapping module, which can internally include a "SLAM Mapping" sub-module and a "Semantic Recognition" sub-module, is responsible for this process. This module first uses the sensors carried by the robot (preferably an RGBD camera that can provide depth information and color information, but there are no special restrictions on the specific model) to perceive the environment, and executes the simultaneous localization and mapping algorithm through the SLAM mapping sub-module to generate a geometric map of the environment (such as a grid map). At the same time, the semantic recognition sub-module uses the visual data containing spatial coordinates obtained by the RGBD camera to identify the objects in the environment. For the object categories that already exist in the preset semantic model library of the system, the semantic recognition sub-module can perform real-time and automatic recognition, and mark the semantic labels (names, categories), as well as their precise spatial positions and postures, of these objects on the generated geometric map. To improve the recognition efficiency, this semantic recognition process can utilize a graphics processing unit (GPU) for parallel computing acceleration. For new or special semantic objects that are not covered in the semantic model library, or for scenarios that require users to perform personalized definitions (such as Figure 5 shown in the "Personalized Semantics" process), users can use the semantic data management interface provided by the system (usually implemented by the map editing and processing module) through the WEB user interface connected to the "cloud service" to perform manual semantic annotation on the map. This includes adding new object or area semantic annotations at specified positions and specifying attributes such as names and categories for them.

[0043] Unified Map Storage Unit: As Figure 1 and Figure 5 shown, all the constructed and acquired map data, whether geometric information or semantic information, is centrally stored and managed by the unified map storage unit. In the ​ architecture, this can be reflected as the "Map Storage Unit" within the "Map Service". This storage unit preferably uses a relational database (such as SQLite, as described in the technology disclosure document) or a NoSQL database suitable for storing structured and semi-structured data to achieve efficient data retrieval, update, and transaction management. Semantic information, such as the name, position, size, orientation, navigation recommendation points, category of an object, the name, center point, color value, segmented image data of a room or area, and user-defined virtual obstacle information, etc. (such as ​As shown, it is usually serialized in a standard format such as JSON and stored associatively with the corresponding geometric map data (such as raster data and its metadata). In particular, regarding the generation and usage logic of navigation recommendation points: after the semantic mapping module completes the recognition and positioning of an object, the system will automatically calculate and generate a navigation recommendation point for the object. This recommendation point is automatically generated based on the size, category of the object, and preset rules (for example, it is defaultly located 0.8 meters directly in front of the object's orientation). For different categories of objects, this preset distance can be different (for example, 0.8 meters for a refrigerator and 0.4 meters for a charging station). Importantly, the distance value of this automatically generated navigation recommendation point also supports subsequent custom modification by the user through the interaction interface to adapt to specific robot task requirements or environmental layouts. When navigating, if the target is an object, the system will preferentially use this unique and precisely calculated navigation recommendation point as the final navigation target pose.

[0044] Custom Semantic Extension and Management Mechanism: To adapt to the specific requirements of different application scenarios, this system provides a powerful custom semantic extension mechanism. Users can define new semantic categories or connect to the "Personalized Semantics" definition function through the API interface or the supporting user interface (such as ​ as shown, connecting to the WEB side to the "Personalized Semantics" definition function. For example, users can define new object or area categories such as "loading point" and "quality inspection table" for a specific factory environment. When the user defines a new semantic category and attempts to save it, the system will perform validity verification, such as checking whether the newly defined category name conflicts with existing category names. If a name conflict occurs, the system will prompt the user and prevent the save, thus ensuring the uniqueness of the semantic category definition and the consistency of the internal data of the system. For categories defined by the user and not existing in the built-in semantic model library of the system, the system will assign a unified internal identifier to it (for example, using a specific tag value such as 255 in the map data to represent such custom categories) for internal processing and identification by the system. All semantic data, whether generated by automatic scanning and recognition through sensors, or custom-added or modified by the user through the interface, as well as the virtual obstacle information set by the user, will be uniformly stored in the database. Each piece of semantic data contains its position in the map image and its mapping pose information in the real physical world, and each has its corresponding label and name for easy query, understanding, and application by each module of the system. Users can conveniently perform addition, deletion, modification, and query (CRUD) operations on these stored semantic information through the semantic data management interface. For example, the user can select the "sofa" icon on the map and modify its semantic category to "refrigerator" through the drop-down list, and the corresponding icon and attributes will be updated accordingly.

[0045] Semantic Navigation Processing Unit and Its Intelligent Navigation: Combining​ , the semantic navigation processing unit is responsible for parsing and executing semantic-based navigation tasks. Its workflow is as ​ follows: 1. The user or upper-layer application issues a semantic navigation instruction, which usually contains the semantic name of the target, such as "go to Conference Room 1" or "find the nearest charging pile". 2. After receiving the instruction, the semantic navigation processing unit first parses the instruction and extracts the key semantic target name or intention. 3. This unit will query the semantic information of the currently loaded map in the unified map storage unit, and according to the parsed semantic target name, search for and match the corresponding semantic elements (objects or regions) to obtain their precise geographical locations (for example, the navigation recommended point coordinates and orientations of objects, or the central point coordinates of regions). 4. After determining the specific physical pose of the navigation target, the semantic navigation processing unit will construct a navigation target request containing information such as the target location, pose, and possible task types. 5. This unit sends this navigation target request to the underlying navigation execution mechanism (such as the ROS2 Navigation Stack or a custom motion controller) through the ROS2's Action / Client mechanism or other inter-process communication methods. 6. After receiving the target, the navigation execution mechanism will perform global and local path planning and control the robot to move towards the target according to the planned path. 7. Finally, the robot reaches the target location. This semantic navigation processing unit not only supports navigation based on semantic names but also generally is compatible with traditional navigation instructions based on precise map coordinates (x, y, θ). More importantly, the navigation control module in this system is designed to support plug-in navigation algorithm expansion. This means that general communication interface protocols and algorithm access points are reserved inside the system architecture, allowing developers or users to conveniently integrate and deploy different types of advanced semantic navigation algorithms according to specific application requirements. Users can select and enable different navigation algorithm plug-ins by modifying the system configuration parameter file. This design greatly enhances the technological foresight of the system and its processing ability for complex environment navigation tasks.

[0046] Automated Map Lifecycle Management Unit: Refer to ​ , in order to effectively manage the growing map data and optimize storage resources, this system integrates an automated map lifecycle management unit. Its workflow is as ​Shown as follows: 1. SLAM algorithm constructs a map: The robot constructs a new initial geometric map through the SLAM algorithm. 2. The geometric map construction is completed. 3. Add / recognize semantic information: Add or recognize relevant semantic information in the newly constructed semantic map. 4. Save the semantic map data to the database: The new semantic map is saved to the database of the unified map storage unit. 5. Check the total number of maps: Set an upper threshold for the number of stored maps (for example, set a maximum of 100 maps to be retained). The automated map lifecycle management unit continuously monitors the total number of maps stored in the current system. 6. Threshold check: Determine whether the total number of stored maps is greater than the upper threshold for the number of stored maps. 7. Delete the oldest map (if the threshold is exceeded): If the judgment in step 6 is "yes", that is, when it is detected that the total number of maps exceeds the preset threshold, the management unit will trigger a cleaning mechanism. This mechanism selects and deletes redundant old map data according to a preset policy (for example, by default, based on the timestamp of map creation, automatically identify and delete the oldest one or more maps; or it can be configured to be based on more complex policies such as the usage frequency of the map (LRU - Least Recently Used) or map quality score), and enter step 8. 8. Copy to the latest directory (if the threshold is not exceeded or the oldest map is deleted when the threshold is exceeded) When it is detected that the total number of maps exceeds the preset threshold or during the operation of deleting old map data in step 7, the system can selectively automatically save copies of the newly generated maps or maps marked as important to a quick access directory (such as the "latest" directory) so that users or the system can quickly call the latest valid maps. 9. The map lifecycle management is completed. This mechanism ensures the efficient utilization of system resources, avoids the infinite expansion of storage space, and ensures that the maps used for robot navigation are the latest or most relevant map data.

[0047] Virtual Obstacle Management and Its Applications This system allows users to flexibly set virtual obstacles on an existing map through an interactive user interface. In the map editing mode, users can select the corresponding tools and then draw virtual walls (represented as line segments) or virtual restricted areas on the visual display area of the map by means of mouse dragging. Users can also assign names and types to these virtual obstacles. The defined virtual obstacle information, including its precise sequence of coordinate points and attributes, is regarded as a special semantic information and updated by the map operation module into the currently loaded map data, and synchronized to the database of the unified map storage unit for persistent storage. When the robot executes a navigation task, the semantic navigation processing unit reads this virtual obstacle information from the map data during path planning. This information is published to the path planner of the navigation system (for example, the geometry and position of the virtual obstacle are published through ROS topics). The cost map layer or other relevant modules in the navigation system subscribe to this information and integrate these virtual obstacle information into the navigation cost map, considering them as high-cost areas or completely impassable areas. Therefore, the path planning algorithm automatically avoids these user-defined virtual walls and restricted areas when calculating the optimal path. This function enables users to dynamically adjust the allowable activity range of the robot according to temporary needs or specific tasks without making any changes to the physical environment, greatly enhancing the flexibility and safety of robot navigation.

[0048] Cloud Service Integration and Remote Interaction Capabilities As shown in the system component relationship diagram ​ , this system is designed with an interface to the "cloud service". This cloud service platform can serve as a bridge connecting the robot system with remote users or higher-level management systems. Users can access the user interface (WEB side) deployed on the cloud service through a standard WEB browser to perform a series of remote operations, such as remotely viewing the current status and map information of the robot. Through the "personalized semantics" definition interface, remotely adding or modifying the categories and attributes of specific semantic objects not included in the robot model library. Remotely editing the map, including setting virtual obstacles, managing the annotated semantic information, etc. Remotely issuing navigation tasks or other control instructions to the robot. The system has a unified command receiving interface (i.e., the cloud / user interface), which can process instructions from the cloud or other higher-level management systems and dispatch these instructions to the corresponding processing modules for execution through the system scheduling unit, and can also feedback the status information and task execution results of the robot to the cloud. This cloud integration ability greatly expands the application scenarios of the robot, enabling it to integrate into a wider range of Internet of Things (IoT) and intelligent management systems.

[0049] Those skilled in the art should note that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A robot semantic map management and navigation system, characterized in that, Comprising: A unified map storage unit configured to store map data of the working environment of a robot, the map data including geometric map data and semantic information associated with the geometric map data, the semantic information at least including semantic tags automatically generated by real-time semantic recognition of the robot sensor data and semantic tags custom-added by a user through a user interface; A semantic mapping module configured to perform simultaneous localization and mapping (SLAM) based on the perception data of a robot sensor (such as an RGBD camera) to generate the geometric map information, and integrated with a semantic recognition function, the semantic recognition function being capable of automatically generating or updating the semantic tags of semantic objects existing in a model library in the map in real time; A semantic navigation processing unit connected to the unified map storage unit, configured to query the semantic information in the unified map storage unit according to an input navigation instruction based on the semantic information to determine a navigation target position, and control the robot to execute a navigation task based on the determined navigation target position; An automated map lifecycle management unit connected to the unified map storage unit, configured to monitor the number of maps stored in the unified map storage unit, and automatically manage historical map data according to a preset policy when the number of maps exceeds a preset threshold; A virtual obstacle management unit connected to the map storage module, configured to allow a user to define virtual obstacles, and update and store the virtual obstacles as part of the semantic information in the unified map storage unit to serve as navigation constraints during navigation planning by the semantic navigation processing unit; And A system scheduling unit and multiple independent functional processing modules, the functional processing modules at least including a map operation module, a mapping operation module, and a navigation control module that is part of the semantic navigation processing unit; the system scheduling unit is designed using a strategy pattern, executes corresponding strategies according to operation instructions, and communicates with the multiple independent functional processing modules through a unified interface protocol based on a custom ROS2 message type.

2. The robot semantic map management and navigation system according to claim 1, wherein During the semantic recognition process, the semantic mapping module uses a graphics processing unit (GPU) for acceleration processing.

3. The robot semantic map management and navigation system according to claims 1 to 2, characterized in that, The unified map storage unit specifically uses a database to store map data, and the semantic information is serially stored in association with the geometric map data in JSON format.

4. The robot semantic map management and navigation system according to claims 1 to 3, characterized in that, The semantic information further includes object semantic information, the object semantic information recording the name, position, size, orientation, navigation recommended points, and category of an object in the environment; the navigation recommended points are automatically calculated and generated at a preset or user-modifiable specified distance in the direction of the object during the semantic mapping completion stage.

5. The robot semantic map management and navigation system according to any one of claims 1 to 4, characterized in that When the automated map lifecycle management unit deletes maps that meet a specific policy, the specific policy is to delete the earliest created map based on a timestamp and automatically back up the latest map.

6. The robot semantic map management and navigation system according to any one of claims 1 to 5, characterized in that, The virtual obstacles defined by the virtual obstacle management module include virtual walls, represented as line segments, and / or virtual restricted areas, represented as polygons.

7. The system according to claims 1-6, wherein The unified interface protocol based on the custom ROS2 message type includes specific message files (.msg) defined for map operations, mapping operations, and navigation operations respectively, and includes these message files in a custom service file (.srv) for communication with cloud services.

8. The system according to any one of claims 1-7, characterized in that, It also includes a semantic data management interface unit that allows users to perform addition, deletion, modification, and query operations on the semantic information stored in the unified map storage unit through the user interface; when the user customizes a new semantic category, if the defined name duplicates an existing category name, the system will prompt and prevent saving.

9. The system according to any one of claims 1-8, characterized in that, The navigation control module supports plug-in navigation algorithm extension and can select and configure different navigation algorithms through a configuration parameter file.

10. The system according to claims 1-9, characterized in that, It also includes a cloud service interface for communicating with cloud services, which can be connected to the WEB side and the personalized semantic definition interface.

Images