Autonomous navigation method based on BIM-SLAM technology

Abstract

The invention provides an autonomous navigation method based on a BIM-SLAM technology, relates to the field of autonomous navigation and environmental perception, and solves the problems of positioning drift, misjudgment of closed-loop detection and poor dynamic environmental adaptability of a traditional SLAM system in a complex building environment. The method comprises the steps of performing path simulation based on a BIM model to obtain initial session data in a simulation environment; in the actual environment of the target area, when the robot carries out environment scanning and path execution through the SLAM system, actual session data is correspondingly generated; aligning a preset number of actual session data with the initial session data, performing closed-loop detection, and correcting the drift error of the map; based on the difference between the actual session data and the initial session data, newly added or removed objects in the actual environment of the target area are detected, and the map is updated and output to serve as the autonomous navigation type movement basis of the robot. According to the invention, the navigation precision and robustness of the robot in a complex environment are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous navigation and environmental perception, and is applied to a construction freight robot with a SLAM system, and in particular to an autonomous navigation method based on BIM-SLAM technology. Background Art

[0002] With the rapid development of robotics technology, autonomous navigation and environmental perception have become indispensable functions for indoor robots and unmanned vehicles. Traditionally, such systems mainly rely on simultaneous localization and mapping (SLAM) technology, which uses sensors such as lidar, cameras, and IMU to obtain data about the surrounding environment in real time, gradually build maps, and determine their own positions. However, conventional SLAM methods face many challenges when applied in large and complex building environments.

[0003] In complex buildings, such as shopping malls, airport terminals, hospitals, and office buildings, there are a large number of repeated structures and similar features, such as corridors, rooms, and floor layouts. These characteristics make traditional loop detection difficult. Loop detection is based on visual or geometric features in the environment, but it is easy to make misjudgments when faced with repeated structures, causing positioning drift and affecting the consistency of the map. This problem is further exacerbated in crowded areas in dynamic environments, where the interference of moving objects on sensor data increases the difficulty of loop detection.

[0004] When a robot moves in a larger environment, the errors of the SLAM system will gradually accumulate as the distance traveled increases. If loop closure detection cannot be effectively performed to correct these errors, the accuracy of the map will inevitably decrease. For tasks that require long-term operation and involve multiple scans, this error accumulation will seriously affect the final map quality and positioning accuracy.

[0005] In addition, complex indoor environments are often accompanied by frequent changes, such as newly added partitions, removed furniture, or the layout of temporary facilities. Traditional SLAM systems usually assume that the environment is static, so it is difficult to adapt to these changes. This not only hinders the timely update of the map, but also reduces the practicality and reliability of the system.

[0006] Finally, maps generated by conventional SLAM are mostly in the form of sparse point clouds or grids, lacking semantic information, that is, they cannot clearly identify functional areas such as rooms, corridors, and stairs. In complex building environments, such maps limit the robot's understanding and utilization efficiency of space, thereby affecting the navigation effect and the rationality of path planning. Therefore, in order to improve the robot's autonomous navigation capabilities in complex environments, the above problems need to be solved as soon as possible to achieve more accurate and efficient positioning and map construction. Summary of the invention

[0007] The purpose of the present invention is to solve the problems of positioning drift, closed-loop detection misjudgment and poor adaptability to dynamic environments in traditional SLAM systems in complex building environments. Therefore, an autonomous navigation method based on BIM-SLAM technology is proposed. The introduction of building information model BIM provides new possibilities for SLAM systems. The information contained in the BIM model can provide a reliable environmental reference for the SLAM system, enabling the robot to have higher navigation accuracy and robustness in complex environments; the semantic information provided by the BIM model can also help the robot understand the structural characteristics of the surrounding environment, thereby optimizing path planning and obstacle avoidance capabilities.

[0008] The present invention adopts the following technical solutions to achieve the purpose:

[0009] An autonomous navigation method based on BIM-SLAM technology is applied to a robot equipped with a SLAM system, and the robot performs autonomous navigation movement in a target area; the method comprises the following steps:

[0010] S1. Obtain the BIM model corresponding to the target area, perform path simulation based on the BIM model, and obtain the initial session data in the simulation environment as the basis for the multi-session anchor alignment operation;

[0011] S2. In the actual environment of the target area, every time the robot performs an environment scan and path execution through the SLAM system, an actual session data is generated;

[0012] S3, performing a multi-session anchor alignment operation, aligning a preset number of actual session data with the initial session data, and performing closed-loop detection between different actual session data to correct the drift error of the map in the SLAM system;

[0013] S4. After the multi-session anchor alignment operation is completed, based on the difference between the actual session data and the initial session data, the newly added or removed objects in the actual environment of the target area are detected, the map in the SLAM system is updated and output, and the updated map is used as the basis for the robot's autonomous navigation movement.

[0014] In step S1, the present invention obtains initial session data based on the BIM model, which utilizes the geometric structure and semantic information provided by the BIM model, such as walls, doors and windows, room layout, etc., to perform path planning. At the same time, by simplifying and converting the BIM model, the corresponding occupancy grid map OGM is generated, and obstacles and passage areas in the building are marked, thereby providing a basis for path planning and obstacle avoidance. After setting the navigation path points in the simulation environment, the robot's key postures and sensor data are recorded, that is, the initial session data is formed for subsequent multi-session anchor alignment operations.

[0015] In step S2, each actual session data includes a posture graph, which is composed of key frames of the robot at multiple different positions; each key frame includes the robot's posture data, position data and 3D point cloud data.

[0016] In step S3, the present invention uses the reference points (anchor nodes) of the BIM model to align the data of multiple actual sessions to ensure that the data of multiple actual sessions can remain consistent in the global coordinate system. Since the actual sessions are aligned with the BIM model through anchor nodes, the data accumulation error is reduced, and high-precision posture graph optimization can be achieved. The SLAM system can perform closed-loop detection between different actual sessions to match similar environmental features; when a closed loop is detected, the system can align the current session data with the historical data, correct the drift error in the map, and ensure the overall consistency and accuracy of the map. After further adopting the posture graph optimization algorithm, the navigation accuracy of the system in complex structural environments can be improved.

[0017] In step S4, in order to adapt to the dynamic changes of the environment, the present invention introduces a change detection function based on the BIM model in the SLAM process. That is, through the difference between the actual session data and the initial session data, the objects added or removed in the actual environment of the target area are detected and marked as "positive differences"; for these difference data, cluster analysis and surface reconstruction are further performed to incorporate these changes into the current map to ensure that the map can be dynamically updated. This step ensures that the robot can still accurately locate in a dynamic environment and maintain the real-time nature of the map, thereby improving the adaptability of the system in complex building scenes.

[0018] In summary, due to the adoption of this technical solution, the beneficial effects of the present invention are as follows:

[0019] The present invention deeply integrates the BIM model with the SLAM system, which can not only build a high-precision environmental map for the robot, but also use the architectural semantic information provided by the BIM model to achieve more intelligent navigation and real-time map update effects. BIM-SLAM technology greatly improves the accuracy and robustness of autonomous navigation in complex building environments, providing a solid foundation for the robot navigation system.

[0020] In the early stage of navigation, the present invention can use the BIM model simulation method to quickly build an initial map and reduce the robot's exploration time of unknown environments. Since the data of the BIM model contains detailed architectural design parameters, such as the specific positions of walls, doors, windows and other elements, the robot can perform initial positioning more accurately and optimize its posture estimation process. Compared with traditional SLAM technology, this method significantly reduces the accumulation of positioning errors and ensures high-precision navigation performance during long-term operation, which is of great benefit to unmanned freight vehicles traveling through buildings during the construction phase.

[0021] In order to cope with the complex closed-loop detection challenges inside large buildings, this paper introduces a closed-loop detection mechanism based on the BIM model. This mechanism not only improves the accuracy of closed-loop detection, but also reduces the cases of misjudgment and missed judgment. Specifically, the semantic information provided by the BIM model helps to distinguish similar structures, enabling the system to reliably identify loop paths in large-scale building environments, further enhancing the consistency and stability of the map.

[0022] In the face of dynamically changing indoor environments, especially during construction and decoration stages, the present invention has change detection and map update functions. These features allow the autonomous navigation system to reflect changes in the environment in real time, such as newly added or removed furniture, temporarily arranged facilities, etc. By updating the map in a timely manner, the robot can still maintain efficient and accurate navigation capabilities in the face of a constantly changing environment. In addition, this instant map update capability is particularly critical for long-term mission execution, as it ensures the timeliness and accuracy of the map, thereby supporting more reliable autonomous navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a general diagram of the overall process of the autonomous navigation method of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0025] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Example

[0027] An autonomous navigation method based on BIM-SLAM technology. Figure 1 The overall process of the method is shown, which can be viewed simultaneously. The various steps of the method are summarized as follows:

[0028] S1. Obtain the BIM model corresponding to the target area, perform path simulation based on the BIM model, and obtain the initial session data in the simulation environment as the basis for the multi-session anchor alignment operation;

[0029] S2. In the actual environment of the target area, every time the robot performs an environment scan and path execution through the SLAM system, an actual session data is generated;

[0030] S3, performing a multi-session anchor alignment operation, aligning a preset number of actual session data with the initial session data, and performing closed-loop detection between different actual session data to correct the drift error of the map in the SLAM system;

[0031] S4. After the multi-session anchor alignment operation is completed, based on the difference between the actual session data and the initial session data, the newly added or removed objects in the actual environment of the target area are detected, the map in the SLAM system is updated and output, and the updated map is used as the basis for the robot's autonomous navigation movement.

[0032] This embodiment will introduce the details of each step in detail based on the above step sequence. In general, this embodiment combines the BIM model with the 3D point cloud data (3D LiDAR) involved in the SLAM technology to perform the registration process and build a consistent map for long-term map management of robots in a GPS-free environment.

[0033] First, for the initial session data generation part based on the BIM model in step S1, the data preprocessing of the BIM model is performed. The BIM model contains detailed information of the buildings in the target area, including walls, doors, windows, floors, etc. In order to meet the real-time requirements of the SLAM system, this embodiment simplifies the BIM model and only retains the navigation-related parts of the building information corresponding to the BIM model, such as walls, corridors, room boundaries, etc.

[0034] The BIM model also needs to be converted into a format supported by the SLAM system, such as a 2D plan, a 3D mesh model, or key geometric information extracted from it, so as to align it with the real-time sensor data of the robot. Therefore, in step S1, during the path simulation process, the robot completes the path planning and obstacle avoidance in the simulation environment based on the geometric structure and semantic information of the target area building provided by the BIM model.

[0035] After simplifying and converting the BIM model, this embodiment generates a corresponding occupancy grid map OGM. OGM marks obstacles (such as walls) and traffic areas (such as corridors and rooms) in the building, thereby providing a basis for path planning and obstacle avoidance. OGM can perform path simulation in a simulation environment to ensure that the robot can navigate to the specified location efficiently and safely.

[0036] In the simulation environment, the initial path planning goal is set according to the OGM, and after the corresponding navigation path points are divided, the path simulation process is performed by assuming that the robot moves along the navigation path points. During the path simulation process, the posture data, position data, and 3D point cloud data of the robot at each navigation path point are recorded. These three data together constitute the initial session data of the robot in the simulation environment. Since the initial session data records the key posture of the robot and the environmental feature information of the target area, it provides basic support for subsequent multi-session anchor alignment and closed-loop detection.

[0037] In step S2, each actual session data includes a posture graph, which is composed of key frames of the robot at multiple different positions; each key frame includes the robot's posture data, position data and 3D point cloud data.

[0038] In step S3, the multi-session anchor alignment operation is performed based on the anchor node preset in the initial session data; the anchor node is a reference point selected in the BIM model, and the coordinates of the reference point are used as the common coordinates of the actual session data and the initial session data, so that the actual session data and the initial session data are aligned to a unified global navigation coordinate system. This embodiment recommends that the anchor node be selected at the entrance of the building or other specific landmark locations, so that all session data are aligned based on the anchor node, thereby ensuring the uniformity of the data and reducing cumulative errors.

[0039] Similarly, in step S3, each actual session data is based on the initial session data corresponding to the BIM model, and introduces feature descriptors, such as "polar coordinate context", to help match similar locations in different sessions. Similar locations are matched in different actual session data through feature descriptors, so as to perform closed loop detection between different actual session data.

[0040] In this embodiment, between different actual session data, the SLAM system performs closed loop detection by matching feature descriptors and searching the robot's location area to determine whether the robot has returned to the previously visited area during the autonomous navigation movement, thereby determining whether the loop is closed. After the closed loop is detected and determined, the current actual session data is matched with the remaining actual session data in the historical records, thereby correcting and updating the drift error in the map and maintaining the continuity of the map.

[0041] Since the closed-loop detection process of this embodiment introduces the architectural semantic information in the BIM model, the process is more reliable, especially in complex environments with continuous repetitive structures. Since the system has pre-acquired the BIM model, the misjudgment of closed-loop detection is reduced. During the navigation process, the robot can be more clear about whether it has returned to a position it has been to, rather than making misjudgments in multiple repetitive structures.

[0042] After the closed-loop detection is completed in this embodiment, the SLAM system uses a posture graph optimization algorithm to globally optimize the posture graph of the actual session data, so that the actual session data is uniformly aligned in the global map of the SLAM system, thereby updating the global map. The posture graph optimization algorithm used here can be the iSAM2 algorithm; during the global optimization process, the posture offset in the actual session data is corrected and matched to the unified global navigation coordinate system, and the anchor node coordinates are updated at the same time, and the corresponding constraints after the closed loop is determined are added to complete the global optimization, effectively improving the map accuracy.

[0043] In step S4, after the multi-session anchor alignment operation is completed, the SLAM system performs a positive difference detection process. Specifically, the current actual session data is compared with the initial session data obtained based on the BIM model, or with the remaining actual session data in the historical records, to detect objects or areas that are added or removed in the actual environment of the target area. The difference data detected during the comparison process is marked as a positive difference, such as newly added walls, furniture and other objects. This detection process helps the SLAM system identify dynamic changes in the environment and ensure that the map is always consistent with the current environment.

[0044] Subsequently, cluster analysis is performed on the 3D point cloud data corresponding to the positive difference, so that the positive differences are grouped and formed into independent objects or regions. The clustered data is used to reconstruct the global map in the SLAM system, incorporating the newly added objects or regions into the original global map, or deleting the removed objects or regions from the original global map, and then generating an updated surface model, which is output as a new global map in the SLAM system. The map is thus dynamically updated and smoothly reflects changes in the environment.

[0045] After completing the above-mentioned method steps of the present embodiment, the SLAM system will output a highly consistent and high-precision global map for the autonomous navigation movement of the robot in the target area. The map not only contains the static building structure (such as walls, room layout) of the current environment, but also dynamically reflects the changes in the environment (such as newly added or removed objects), ensuring that the robot can navigate based on the latest environmental information in future tasks. When the robot is running continuously, after fully executing the full steps of the method once, it can be cycled between the processes of step S2 to step S4, maintaining efficient autonomous navigation movement, and completing corresponding work tasks, such as building materials transportation.

[0046] By referring to the BIM model, this embodiment can more reliably identify the situation of returning to the explored area, reducing the problem of misjudgment in the repeated structure environment. Unlike traditional SLAM technology that mainly relies on visual or geometric features for closed-loop detection, this embodiment introduces the architectural semantic information in the BIM model for closed-loop detection. This method significantly enhances the robustness of closed-loop detection, especially in complex or repeated structural environments in large buildings.

[0047] In summary, this embodiment provides an innovative autonomous navigation solution for complex building environments. It not only improves the accuracy and reliability of the navigation system, but also proposes effective strategies for long-term map management and dynamic environment adaptation, greatly expanding the potential of robotics technology in practical applications.

Claims

1. An autonomous navigation method based on BIM-SLAM technology, characterized in that: The method is applied to a robot equipped with a SLAM system, and the robot performs autonomous navigation movement in a target area; the method comprises the following steps: S1. Obtain the BIM model corresponding to the target area, perform path simulation based on the BIM model, and obtain the initial session data in the simulation environment as the basis for the multi-session anchor alignment operation; S2. In the actual environment of the target area, every time the robot performs an environment scan and path execution through the SLAM system, an actual session data is generated; S3, performing a multi-session anchor alignment operation, aligning a preset number of actual session data with the initial session data, and performing closed-loop detection between different actual session data to correct the drift error of the map in the SLAM system; S4. After the multi-session anchor alignment operation is completed, based on the difference between the actual session data and the initial session data, the newly added or removed objects in the actual environment of the target area are detected, the map in the SLAM system is updated and output, and the updated map is used as the basis for the robot's autonomous navigation movement.

2. The autonomous navigation method according to claim 1, characterized in that: During the path simulation process of step S1, the robot completes path planning and obstacle avoidance in the simulation environment based on the geometric structure and semantic information of the target area buildings provided by the BIM model.

3. The autonomous navigation method according to claim 2, characterized in that: After simplifying and converting the BIM model, only the navigation-related part of the building information corresponding to the BIM model is retained to generate the corresponding occupancy grid map OGM; after marking the obstacles and passage areas in the target area in the OGM, it provides a basis for path planning and obstacle avoidance.

4. The autonomous navigation method according to claim 3, characterized in that: In the simulation environment, the initial path planning goal is set according to the OGM, and after dividing the corresponding navigation path points, the path simulation process is performed by assuming that the robot moves along the navigation path points; during the path simulation process, the robot's posture data, position data and 3D point cloud data at each navigation path point are recorded. These three data together constitute the robot's initial session data in the simulation environment.

5. The autonomous navigation method according to claim 1, characterized in that: In step S2, each actual session data includes a posture graph, which is composed of key frames of the robot at multiple different positions; each key frame includes the robot's posture data, position data and 3D point cloud data.

6. The autonomous navigation method according to claim 5, characterized in that: In step S3, the multi-session anchor alignment operation is performed based on the preset anchor node in the initial session data; the anchor node is a reference point selected in the BIM model, and the coordinates of the reference point are used as the common coordinates of the actual session data and the initial session data, so that the actual session data and the initial session data are aligned to a unified global navigation coordinate system.

7. The autonomous navigation method according to claim 6, characterized in that: In step S3, each actual session data is based on the initial session data corresponding to the BIM model and introduces a feature descriptor; By matching similar locations in different actual session data with feature descriptors, closed-loop detection is performed between different actual session data.

8. The autonomous navigation method according to claim 7, characterized in that: Between different actual session data, the SLAM system performs closed-loop detection by matching feature descriptors and searching the area where the robot is located, to determine whether the robot has returned to an area it has visited before during autonomous navigation movement, thereby determining whether the loop is closed; after detecting and determining a closed loop, the current actual session data is matched with the rest of the actual session data in the historical records, thereby correcting and updating the drift error in the map and maintaining the continuity of the map.

9. The autonomous navigation method according to claim 8, characterized in that: After the closed-loop detection is completed, the SLAM system uses a posture graph optimization algorithm to globally optimize the posture graph of the actual session data, so that the actual session data is uniformly aligned in the global map of the SLAM system, thereby updating the global map; During the global optimization process, the pose offset in the actual session data is corrected and matched to the unified global navigation coordinate system. At the same time, the anchor node coordinates are updated, and the corresponding constraints after the closed loop are added to complete the global optimization.

10. The autonomous navigation method according to claim 1, characterized in that: In step S4, after the multi-session anchor alignment operation is completed, the SLAM system performs a positive difference detection process; the current actual session data is compared with the initial session data obtained based on the BIM model, or with the remaining actual session data in the historical records, to detect objects or areas that are added or removed in the actual environment of the target area; After marking the difference data detected in the comparison process as positive differences, cluster analysis is performed on the 3D point cloud data corresponding to the positive differences, so as to group the positive differences and form independent objects or regions; The clustered data is used to reconstruct the global map in the SLAM system, incorporating the newly added objects or areas into the original global map, or deleting the removed objects or areas from the original global map, thereby generating an updated surface model and outputting it as a new global map in the SLAM system; the robot continues to perform autonomous navigation movement in the target area based on the updated global map, and the SLAM system continues to execute the loop process from step S2 to step S4 during the movement.

Images