Large-space laser navigation precision measuring and calculating method based on SLAM (Simultaneous Localization and Mapping) technology

Abstract

A large-space laser navigation precision measuring and calculating method based on an SLAM technology belongs to the field of laser navigation and comprises the following steps: step 1, acquiring environmental data by using a laser radar; 2, constructing a three-dimensional map of a large-space environment according to environment data acquired by the laser radar; 3, robot walking path planning or robot walking area setting is carried out on the three-dimensional map constructed in the step 2; 4, in the walking process of the robot, data and map information are obtained through the laser radar, and positioning operation on the robot is completed; 5, continuously updating and constructing a surrounding environment map of the robot in the moving process of the robot; and step 6, verifying the fault object. According to the invention, the precision and consistency of three-dimensional map construction are improved by optimizing the laser radar data acquisition and preprocessing process and combining the feature point extraction and loopback detection mechanism; and meanwhile, a multi-factor fusion real-time precision evaluation model is introduced, so that the navigation state of the robot is dynamically monitored and fed back.

Description

Technical Field

[0001] This invention belongs to the field of laser navigation, specifically a method for measuring the accuracy of large-space laser navigation based on SLAM technology. Background Technology

[0002] Laser navigation technology utilizes laser sensors to achieve 3D modeling and localization of the environment, enabling precise navigation and obstacle avoidance for robots. Its main purpose is to improve robots' navigation capabilities in both indoor and outdoor environments, allowing them to accurately identify and understand their surroundings, and to plan and execute paths. Through laser navigation technology, robots can effectively avoid obstacles and hazards, and complete various tasks.

[0003] Traditional laser navigation methods often have limitations in adaptability to large-space environments. On the one hand, the sparsity or repetition of environmental features in large-space scenarios can easily lead to accumulated errors in SLAM algorithms during localization. Especially in long-distance movement or complex dynamic environments, the localization accuracy will gradually decrease over time, making it difficult to meet the requirements of high-precision navigation. On the other hand, the real-time evaluation mechanism for navigation accuracy in existing technologies is not perfect. It usually judges accuracy only by comparing the deviation between the actual trajectory and the planned trajectory after the fact, and cannot dynamically feed back the current accuracy status during robot operation. This causes the robot to continue to execute the original path when the accuracy is insufficient, which may lead to collision risks or task failure. In addition, when facing sudden obstacles, the map update efficiency and fault identification accuracy of traditional methods need to be improved. In particular, for dynamic obstacles that do not exist in the 3D map in advance, the response speed of position marking and path replanning is slow, which affects the robot's autonomous operation capability in complex environments. Summary of the Invention

[0004] This invention provides a method for measuring the accuracy of large-space laser navigation based on SLAM technology, in order to overcome the shortcomings of existing technologies.

[0005] This invention is achieved through the following technical solution: The method for calculating the accuracy of large-space laser navigation based on SLAM technology includes the following steps: Step 1: Acquire environmental data using lidar; Step 2: Construct a 3D map of the large-space environment based on the environmental data acquired by the lidar; Step 3: Plan the robot's walking path or set the robot's walking area on the 3D map constructed in Step 2; Step 4: During the robot's movement, it uses LiDAR to acquire data and map information to complete the robot's localization operation; Step 5: During the movement, the robot obtains the surrounding situation in real time through LiDAR. The robot continuously updates and builds a map of its surrounding environment. If there are obstacles on the movement path, the robot automatically replans the route and marks the location of the faulty object on the 3D map. Step Six: Verify the faulty object, determine the relevant information of the faulty object that does not exist in the 3D map, and update the 3D map in real time.

[0006] As described above, in the large-space laser navigation accuracy measurement method based on SLAM technology, in step one, the laser radar is moved within a large space by a robot to collect environmental information. During the movement, the laser radar continuously scans the surrounding environment at a frequency of 100Hz, with a single scan containing no less than 10,000 point clouds. The scanning angle covers a horizontal range of 360° and a vertical range of ±30°, ensuring comprehensive capture of environmental features at different heights within the large space. This guarantees the density and continuity of environmental data, meeting the requirements for detail accuracy in 3D map construction.

[0007] As described above, the large-space laser navigation accuracy measurement method based on SLAM technology involves the following steps in the second step: First, the raw point cloud data is preprocessed, including noise reduction, filtering, and coordinate calibration. A statistical filtering algorithm is used to calculate the distance distribution between each point and its neighboring points, eliminating outliers with a mean distance deviation exceeding three times the standard deviation. This reduces the data volume and improves subsequent processing efficiency while preserving the environmental feature contours. Coordinate calibration then uses data from the robot's built-in IMU sensor to perform spatiotemporal synchronization correction on the point cloud coordinates collected by the laser radar, ensuring that data collected at different times are in a unified coordinate system.

[0008] As described above, in the large-space laser navigation accuracy measurement method based on SLAM technology, after the preprocessing in step two, a 3D map is constructed using a feature-point-based SLAM algorithm. Stable key feature points are extracted from the point cloud map, feature point matching is performed, and point cloud registration is completed. This results in the construction of a 3D point cloud map that includes environmental topology and geometric features. At the same time, a loop closure detection mechanism is introduced during the map construction process to identify scenes where the robot arrives at the same location at different times, calculate the cumulative error, and perform global optimization to improve the overall consistency and accuracy of the 3D map.

[0009] As described above, in the large-space laser navigation accuracy calculation method based on SLAM technology, the path planning in step three comprehensively considers path length, obstacle distribution, and navigation accuracy requirements to generate the robot's walking trajectory; the area setting can delineate specific walking area boundaries in the three-dimensional map according to the spatial range parameters input by the user, ensuring that the robot moves within the preset range and avoids exceeding the task execution space.

[0010] As described above, in the large-space laser navigation accuracy calculation method based on SLAM technology, in step four, the robot matches the real-time environmental data collected by the LiDAR with the 3D map constructed in step two during its movement. By comparing feature points and transforming coordinates, the robot's precise position coordinates in the 3D map are calculated. Simultaneously, the robot's own odometry data is combined for fusion positioning. The LiDAR positioning result and the odometry positioning result are fused to reduce the impact of single sensor errors, further improve positioning accuracy, and ensure that the robot's position information is accurate and reliable when moving in a large space.

[0011] As described above, in the large-space laser navigation accuracy measurement method based on SLAM technology, step four involves extracting inter-frame matching uncertainty factors, environmental feature degradation factors, loop closure detection reliability factors, and global trajectory consistency factors from within the laser SLAM system in real time during robot operation. The inter-frame matching uncertainty factor is calculated based on the matching residuals and covariance matrices of the iterative nearest-point algorithm during the laser frame and local map matching process, determining translational and rotational uncertainties. The environmental feature degradation factor quantifies the richness and degradation direction of environmental features by analyzing the geometric distribution characteristics of the laser scanning data and calculating eigenvalues ​​or flatness. The loop closure detection reliability factor assesses the reliability of the loop when loop closure occurs, including matching score, historical consistency check, and the residual magnitude of the loop constraint in pose graph optimization. Simultaneously, the extracted inter-frame matching uncertainty factors, loop closure detection reliability factors, and global trajectory consistency factors are adaptively weighted and fused using an accuracy estimation model.

[0012] The accuracy estimation model of the large-space laser navigation accuracy measurement method based on SLAM technology, as described above, is calculated using the following formula: in, This represents the overall navigation accuracy score at time t. This represents the adaptive weight of the i-th precision factor. This represents the normalized value of the i-th precision factor. The value range is [0,1]. The closer the value is to 1, the higher the accuracy, and the closer it is to 0, the lower the accuracy. The inter-frame matching uncertainty factor, environmental feature degradation factor, closed-loop detection reliability factor and global trajectory consistency factor are imported and their expressions are: in, The value of the inter-frame matching uncertainty factor after normalization is the final residual after the ICP algorithm performs inter-frame matching calculation and converges. At the same time, the covariance is approximately estimated based on the Jacobian matrix of ICP to obtain the uncertainty of translation and rotation. The uncertainty value is mapped to the [0,1] interval for normalization. The environmental feature degradation factor is the normalized value, which is calculated by performing principal component analysis on the current frame point cloud, calculating the eigenvalues ​​λ1, λ2, and λ3, and then calculating the entropy value. To measure the divergence of the distribution, the smaller the entropy value, the more concentrated the features are, which may indicate degradation. The entropy value is obtained after normalization. This is the normalized value of the closed-loop detection reliability factor. When a closed loop is detected, the matching score and the degree of consistency with historical trajectories are recorded. If the matching score is high and the historical consistency test passes, then a reliability factor is assigned. A high value, otherwise assign A low value, if no closed loop occurs, Use the default value of 0.5; The global trajectory consistency factor is the normalized value. After optimization, the average residual of all edges is calculated. The smaller the average residual, the better the global trajectory consistency. The higher the value; Set initial weights: =0.3, w2=0.3, w3=0.2, w4=0.2; if the current environment is identified as feature degradation, i.e. If w2 is less than 0.2, temporarily increase w2 to 0.5 and correspondingly reduce other weights to emphasize the main impact of the current environment on accuracy; if a high-reliability closed-loop operation has just occurred... If the value is greater than 0.9, the weight of w3=0.2 will be significantly increased in the subsequent period because the closed loop has a decisive impact on the global accuracy. Preset Threshold, when When the value is greater than 0.8, the robot runs at full speed; when the value is less than 0.5... When the velocity is ≤0.8, the robot slows down and attempts repositioning; when... When the value is ≤0.5, the robot immediately stops and issues a warning.

[0013] As described above, in the large-space laser navigation accuracy measurement method based on SLAM technology, step five involves the robot updating data on areas within its current field of view that have changed, in order to reduce computational resource consumption.

[0014] As described above, in the large-space laser navigation accuracy measurement method based on SLAM technology, step six involves identifying obstacles on the movement path. By comparing the differences between the real-time acquired data and the corresponding area in the original 3D map, and combining the obstacle's size, shape, and movement trend, the system determines whether the obstacle is a static or dynamic obstacle. For static obstacles, their specific location coordinates and outline information are directly marked on the 3D map. For dynamic obstacles, in addition to marking the current location, their movement trajectory data is also recorded for reference during subsequent path planning. If the replanned route deviates from the original route by more than a preset threshold, the system will automatically perform a secondary verification of the feasibility of the new route to ensure that the replanned route is safe and efficient.

[0015] The advantages of this invention are: by optimizing the LiDAR data acquisition and preprocessing process, and combining feature point extraction and loop closure detection mechanisms, the accuracy and consistency of 3D map construction are improved; at the same time, a multi-factor fusion real-time accuracy evaluation model is introduced to realize dynamic monitoring and feedback of the robot's navigation status; in the obstacle recognition and path planning stages, the classification and processing of static and dynamic obstacles and the secondary verification mechanism ensure the safety and efficiency of navigation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, the method for calculating the accuracy of large-space laser navigation based on SLAM technology includes the following steps: Step 1: Use LiDAR to acquire environmental data (LiDAR emits laser beams and receives the reflected laser beams, which can provide high-precision environmental data, including the location, shape, and distance of obstacles). Step 2: Construct a 3D map of the large-space environment based on the environmental data acquired by the lidar; Step 3: Plan the robot's walking path or set the robot's walking area on the 3D map constructed in Step 2; Step 4: During the robot's movement, it uses LiDAR to acquire data and map information to complete the robot's localization operation; Step 5: During the movement, the robot obtains the surrounding situation in real time through LiDAR. The robot continuously updates and builds a map of its surrounding environment. If there are obstacles on the movement path, the robot automatically replans the route and marks the location of the faulty object on the 3D map. Step Six: Verify the faulty object, determine the relevant information of the faulty object that does not exist in the 3D map, and update the 3D map in real time.

[0020] Specifically, in step one of this embodiment, the LiDAR is moved within a large space by a robot to collect environmental information. During the movement, the LiDAR continuously scans the surrounding environment at a frequency of 100Hz, with a single scan containing no less than 10,000 point clouds. The scanning angle covers a horizontal range of 360° and a vertical range of ±30°, ensuring comprehensive capture of environmental features at different heights within the large space. This guarantees the density and continuity of environmental data, meeting the requirements for detail accuracy in 3D map construction.

[0021] Specifically, the processing operation of the environmental data acquired by the LiDAR in step two of this embodiment is as follows: First, the original point cloud data is preprocessed, including noise reduction, filtering, and coordinate calibration. A statistical filtering algorithm is used to calculate the distance distribution between each point and its neighboring points, and outliers with a distance mean deviation exceeding three times the standard deviation are removed. This reduces the amount of data and improves the efficiency of subsequent processing while preserving the environmental feature contours. The coordinate calibration is performed by using the IMU sensor data built into the robot to perform spatiotemporal synchronization correction on the point cloud coordinates collected by the LiDAR, ensuring that the data collected at different times are in a unified coordinate system.

[0022] More specifically, after the preprocessing in step two of this embodiment is completed, a 3D map is constructed using a feature-point-based SLAM algorithm. Stable key feature points are extracted from the point cloud map, feature point matching is performed, point cloud registration is completed, and a 3D point cloud map containing environmental topology and geometric features is constructed. At the same time, a loop closure detection mechanism is introduced during the map construction process to identify scenes where the robot arrives at the same location at different times, calculate the cumulative error and perform global optimization to improve the overall consistency and accuracy of the 3D map.

[0023] More specifically, in step three of this embodiment, path planning comprehensively considers path length, obstacle distribution, and navigation accuracy requirements to generate the robot's walking trajectory; area setting can delineate specific walking area boundaries on a 3D map based on the spatial range parameters input by the user, ensuring that the robot moves within a preset range and avoids exceeding the task execution space.

[0024] More specifically, in step four of this embodiment, during the robot's movement, the real-time environmental data collected by the LiDAR is matched with the 3D map constructed in step two. The precise position coordinates of the robot in the 3D map are calculated through feature point comparison and coordinate transformation. At the same time, the robot's own odometry data is combined for fusion positioning. The LiDAR positioning result and the odometry positioning result are fused to reduce the impact of single sensor error, further improve positioning accuracy, and ensure that the robot's position information is accurate and reliable when moving in a large space.

[0025] Furthermore, in step four of this embodiment, during robot operation, inter-frame matching uncertainty factors, environmental feature degradation factors, loop closure detection reliability factors, and global trajectory consistency factors are extracted in real time from within the laser SLAM system. The inter-frame matching uncertainty factor is calculated based on the matching residuals and covariance matrices of the iterative nearest point algorithm during the laser frame and local map matching process, calculating translational and rotational uncertainties. The environmental feature degradation factor is based on analyzing the geometric distribution characteristics of the laser scanning data, quantifying the richness and degradation direction of environmental features by calculating eigenvalues ​​or flatness (curvature). The loop closure detection reliability factor evaluates the reliability of the loop when loop closure occurs, including matching score, historical consistency check, and the residual magnitude of the loop constraint in pose graph optimization. Simultaneously, the extracted inter-frame matching uncertainty factor, loop closure detection reliability factor, and global trajectory consistency factor are adaptively weighted and fused using an accuracy estimation model.

[0026] Furthermore, the calculation formula for the accuracy estimation model described in this embodiment is as follows: in, This represents the overall navigation accuracy score at time t. This represents the adaptive weight of the i-th precision factor. This represents the normalized value of the i-th precision factor. The value range is [0,1]. The closer the value is to 1, the higher the accuracy, and the closer it is to 0, the lower the accuracy. The inter-frame matching uncertainty factor, environmental feature degradation factor, closed-loop detection reliability factor and global trajectory consistency factor are imported and their expressions are: in, The value of the inter-frame matching uncertainty factor after normalization is the final residual after the ICP algorithm performs inter-frame matching calculation and converges. At the same time, the covariance is approximately estimated based on the Jacobian matrix of ICP to obtain the uncertainty of translation and rotation. The uncertainty value is mapped to the [0,1] interval for normalization. The environmental feature degradation factor is the normalized value, which is calculated by performing principal component analysis on the current frame point cloud, calculating eigenvalues ​​λ1, λ2, λ3 (λ1≥λ2≥λ3), and then calculating the entropy value. To measure the divergence of the distribution, the smaller the entropy value, the more concentrated the features are, which may indicate degradation. The entropy value is obtained after normalization. This is the normalized value of the closed-loop detection reliability factor. When a closed loop is detected, the matching score and the degree of consistency with historical trajectories are recorded. If the matching score is high and the historical consistency test passes, then a reliability factor is assigned. A high value (close to 1), otherwise assign A low value, if no closed loop occurs, Use the default value of 0.5; The global trajectory consistency factor is the normalized value. After optimization, the average residual of all edges is calculated. The smaller the average residual, the better the global trajectory consistency. The higher the value; Set initial weights: =0.3, w2=0.3, w3=0.2, w4=0.2; if the current environment is identified as feature degradation, i.e. If w2 is less than 0.2, temporarily increase w2 to 0.5 and correspondingly reduce other weights to emphasize the main impact of the current environment on accuracy; if a high-reliability closed-loop operation has just occurred... If the value is greater than 0.9, the weight of w3=0.2 will be significantly increased in the subsequent period because the closed loop has a decisive impact on the global accuracy. Preset Threshold, when When the value is greater than 0.8, the robot runs at full speed; when the value is less than 0.5... When the velocity is ≤0.8, the robot slows down and attempts repositioning; when... When the value is ≤0.5, the robot immediately stops and issues a warning.

[0027] Furthermore, in step five of this embodiment, the robot updates the data of areas within its current field of view that have changed, in order to reduce the consumption of computing resources.

[0028] Furthermore, in step six of this embodiment, when identifying obstacles on the movement path, the system compares the differences between the real-time collected data and the corresponding area in the original 3D map, and combines the characteristics of the obstacle such as its size, shape, and movement trend to determine whether the obstacle is a static or dynamic obstacle. For static obstacles, their specific location coordinates and outline information are directly marked on the 3D map. For dynamic obstacles, in addition to marking the current location, their movement trajectory data is also recorded for reference during subsequent path planning. If the replanned route deviates from the original route by more than a preset threshold, the system will automatically perform a secondary verification of the feasibility of the new route to ensure that the replanned route is safe and efficient.

[0029] This invention differs from traditional methods in 20,000m 2 Comparative experiments were conducted in open factory settings to verify the effectiveness of the method. Using traditional methods, the robot's cumulative positioning error reached 0.8m after 30 minutes of continuous movement. In contrast, the present invention, through real-time loop closure detection and global optimization, consistently kept the positioning error within 0.15m. In a complex office area containing 10 dynamic pedestrians, the system's response time to dynamic obstacle recognition was less than 0.3s, and the path replanning success rate reached 98.7%. Furthermore, the secondary verification mechanism improved the average driving efficiency of the new route by 12.3% compared to traditional methods. Moreover, even in low-data scenarios where the LiDAR point cloud density dropped to 5000 points / second, the method still maintained a feature extraction accuracy of over 85%, demonstrating its strong adaptability to the sparse features of large-scale spatial environments.

[0030] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring the accuracy of large-space laser navigation based on SLAM technology, characterized in that: Includes the following steps: Step 1: Acquire environmental data using lidar; Step 2: Construct a 3D map of the large-space environment based on the environmental data acquired by the lidar; Step 3: Plan the robot's walking path or set the robot's walking area on the 3D map constructed in Step 2; Step 4: During the robot's movement, it uses LiDAR to acquire data and map information to complete the robot's localization operation; Step 5: During the movement, the robot obtains the surrounding situation in real time through LiDAR. The robot continuously updates and builds a map of its surrounding environment. If there are obstacles on the movement path, the robot automatically replans the route and marks the location of the faulty object on the 3D map. Step Six: Verify the faulty object, determine the relevant information of the faulty object that does not exist in the 3D map, and update the 3D map in real time.

2. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 1, characterized in that: In step one, the LiDAR is moved within a large space by a robot to collect environmental information. During the movement, the LiDAR continuously scans the surrounding environment at a frequency of 100Hz, with a single scan containing no less than 10,000 point clouds. The scanning angle covers a horizontal range of 360° and a vertical range of ±30°, ensuring comprehensive capture of environmental features at different heights within the large space. This guarantees the density and continuity of environmental data, meeting the requirements for detail accuracy in 3D map construction.

3. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 1, characterized in that: The processing of environmental data acquired by the LiDAR in step two is as follows: First, the raw point cloud data is preprocessed, including noise reduction, filtering, and coordinate calibration. A statistical filtering algorithm is used to calculate the distance distribution between each point and its neighboring points, and outliers with a mean distance deviation exceeding three times the standard deviation are removed. This reduces the amount of data and improves the efficiency of subsequent processing while preserving the environmental feature contours. Coordinate calibration is performed by using the robot's built-in IMU sensor data to perform spatiotemporal synchronization correction on the point cloud coordinates acquired by the LiDAR, ensuring that the data acquired at different times are in a unified coordinate system.

4. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 3, characterized in that: After the preprocessing in step two is completed, a 3D map is constructed using a feature-point-based SLAM algorithm. Stable key feature points are extracted from the point cloud map, feature point matching is performed, point cloud registration is completed, and a 3D point cloud map containing environmental topology and geometric features is constructed. At the same time, a loop closure detection mechanism is introduced during the map construction process to identify scenes where the robot arrives at the same location at different times, calculate the cumulative error and perform global optimization to improve the overall consistency and accuracy of the 3D map.

5. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 1, characterized in that: In step three, path planning comprehensively considers path length, obstacle distribution, and navigation accuracy requirements to generate the robot's walking trajectory; area setting can define specific walking area boundaries on a 3D map based on the spatial range parameters input by the user, ensuring that the robot moves within a preset range and avoids exceeding the task execution space.

6. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 1, characterized in that: In step four, during the robot's movement, the real-time environmental data collected by the LiDAR is matched with the 3D map constructed in step two. The precise position coordinates of the robot in the 3D map are calculated through feature point comparison and coordinate transformation. At the same time, the robot's own odometry data is combined for fusion positioning. The LiDAR positioning result and the odometry positioning result are fused to reduce the impact of single sensor error, further improve positioning accuracy, and ensure that the robot's position information is accurate and reliable when moving in a large space.

7. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 1, characterized in that: In step four, during robot operation, the inter-frame matching uncertainty factor, environmental feature degradation factor, loop closure detection reliability factor, and global trajectory consistency factor are extracted in real time from the laser SLAM system. The inter-frame matching uncertainty factor is calculated based on the matching residual and covariance matrix of the iterative nearest point algorithm in the laser frame and local map matching process, calculating translational and rotational uncertainties. The environmental feature degradation factor is based on the analysis of the geometric distribution characteristics of the laser scanning data, quantifying the richness and degradation direction of environmental features by calculating eigenvalues ​​or flatness. The loop closure detection reliability factor evaluates the reliability of the loop when a loop closure occurs, including the matching score, historical consistency check, and the residual magnitude of the loop constraint in pose graph optimization. Simultaneously, the extracted inter-frame matching uncertainty factor, loop closure detection reliability factor, and global trajectory consistency factor are adaptively weighted and fused through an accuracy estimation model.

8. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 7, characterized in that: The calculation formula for the accuracy estimation model is as follows: in, This represents the overall navigation accuracy score at time t. This represents the adaptive weight of the i-th precision factor. This represents the normalized value of the i-th precision factor. The value range is [0,1]. The closer the value is to 1, the higher the accuracy, and the closer it is to 0, the lower the accuracy. The inter-frame matching uncertainty factor, environmental feature degradation factor, closed-loop detection reliability factor and global trajectory consistency factor are imported and their expressions are: in, The value of the inter-frame matching uncertainty factor after normalization is the final residual after the ICP algorithm performs inter-frame matching calculation and converges. At the same time, the covariance is approximately estimated based on the Jacobian matrix of ICP to obtain the uncertainty of translation and rotation. The uncertainty value is mapped to the [0,1] interval for normalization. The environmental feature degradation factor is the normalized value, which is calculated by performing principal component analysis on the current frame point cloud, calculating the eigenvalues ​​λ1, λ2, and λ3, and then calculating the entropy value. To measure the divergence of the distribution, the smaller the entropy value, the more concentrated the features are, which may indicate degradation. The entropy value is obtained after normalization. This is the normalized value of the closed-loop detection reliability factor. When a closed loop is detected, the matching score and the degree of consistency with historical trajectories are recorded. If the matching score is high and the historical consistency test passes, then a reliability factor is assigned. A high value, otherwise assign A low value, if no closed loop occurs, Use the default value of 0.5; The global trajectory consistency factor is the normalized value. After optimization, the average residual of all edges is calculated. The smaller the average residual, the better the global trajectory consistency. The higher the value; Set initial weights: =0.3, w2=0.3, w3=0.2, w4=0.2; if the current environment is identified as feature degradation, i.e. If w2 is less than 0.2, temporarily increase w2 to 0.5 and correspondingly reduce other weights to emphasize the main impact of the current environment on accuracy; if a high-reliability closed-loop operation has just occurred... If the value is greater than 0.9, the weight of w3=0.2 will be significantly increased in the subsequent period because the closed loop has a decisive impact on the global accuracy. Preset Threshold, when When the value is greater than 0.8, the robot runs at full speed; when the value is less than 0.5... When the velocity is ≤0.8, the robot slows down and attempts repositioning; when... When the value is ≤0.5, the robot immediately stops and issues a warning.

9. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 1, characterized in that: In step five, the robot updates the data for areas within its current field of view that have changed, in order to reduce the consumption of computing resources.

10. The method for calculating the accuracy of large-space laser navigation based on SLAM technology according to claim 1, characterized in that: In step six, when identifying obstacles on the movement path, the system compares the differences between the real-time collected data and the corresponding area in the original 3D map, and combines the obstacle's size, shape, and movement trend to determine whether the obstacle is a static or dynamic obstacle. For static obstacles, their specific location coordinates and outline information are directly marked on the 3D map. For dynamic obstacles, in addition to marking the current location, their movement trajectory data is also recorded for reference during subsequent path planning. If the replanned route deviates from the original route by more than a preset threshold, the system will automatically perform a secondary verification of the feasibility of the new route to ensure that the replanned route is safe and efficient.

Images