An intelligent navigation method applicable to the environment of shelf aisles and logistics areas

Abstract

The present invention relates to the technical field of logistics robot navigation, in particular to an intelligent navigation method applicable to the environment of shelf aisles and logistics areas. Point cloud data collected during the operation of an AGV vehicle is obtained. After noise suppression of the point cloud data, an accumulator is constructed based on the Manhattan orthogonal constraint of the workshop to perform orthogonal cumulative weighted statistics on the search neighborhood of the point cloud data, so as to determine the source landmark point cloud features for the AGV vehicle to extract the shelf aisle. Real-time sensing data of multi-modal sensors is obtained, and a real-time multi-sensor fusion model is constructed based on hierarchical factor graph and belief propagation calculation. The source landmark point cloud features are corrected, registered and optimized by using the real-time multi-sensor fusion model and the optical flow distortion amplitude of the dynamic constraint, so as to provide positioning navigation for the AGV vehicle. The present invention can provide the AGV vehicle with powerful environment perception, positioning and navigation capabilities, so as to realize efficient warehousing operations and logistics task execution.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics robot navigation, and particularly to an intelligent navigation method applicable to the environments of shelf aisles and logistics areas. Background Art

[0002] With the rapid development of fields such as autonomous driving, robot navigation, and environmental perception, multi-modal sensor fusion technology plays an increasingly important role in intelligent systems. Especially in complex dynamic scenarios, traditional single sensors often cannot meet the requirements for high precision, real-time performance, and robustness. Multi-modal sensors (such as lidar, cameras, radars, etc.) can provide the system with richer and more comprehensive environmental information. However, in practical applications, due to problems such as asynchrony, data latency, and noise between sensors, the data fusion of these sensors has become a huge challenge. Especially in dynamic environments, the interference of motion distortion and sensor noise will reduce the accuracy and effectiveness of sensor data, affecting tasks such as final target recognition and path planning.

[0003] Point cloud data, as the main data form output by sensors such as lidar, its accuracy and quality are directly related to the perception ability of the system. Point cloud registration technology, as a core technology in point cloud data processing, is used to synthesize point cloud data obtained from multiple different perspectives into a unified global coordinate system. In complex dynamic environments, point cloud registration faces influences such as motion distortion, sensor noise, and data latency. Therefore, it is necessary to develop an intelligent navigation method that can achieve efficient and robust point cloud registration, especially for combating noise interference, to solve the above problems. Summary of the Invention

[0004] The present invention overcomes the deficiencies of the prior art and provides an intelligent navigation method applicable to the environments of shelf aisles and logistics areas.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] The first aspect of the present invention provides an intelligent navigation method applicable to the environments of shelf aisles and logistics areas, including the following steps:

[0007] S102: Obtain the point cloud data collected by the lidar SLAM operation, identify the Q-table degradation evaluation table through the statistical characteristics degradation of the point cloud data, obtain the maximum degradation evaluation value, and calculate the point cloud noise suppression for the lidar SLAM based on the in-degree interference degree of the maximum degradation evaluation value;

[0008] S104: After suppressing the noise of the point cloud data, construct an accumulator based on the Manhattan orthogonality constraint of the factory building to perform orthogonal cumulative weighted statistics on the search neighborhood of the point cloud data, so as to determine the source landmark point cloud features for extracting the shelf channels by lidar SLAM;

[0009] S106: Obtain the real-time sensing data of the multi-modal sensor, construct a real-time multi-sensor fusion model based on the hierarchical factor graph and belief propagation calculation, and use the real-time multi-sensor fusion model and the optical flow distortion amplitude of the dynamic constraint to correct and register and optimize the source landmark point cloud features to provide positioning and navigation for the AGV;

[0010] S108: Continuously incrementally learn and update the tree structure of the initial incremental regression tree of the point cloud data according to the operation requirements of the AGV vehicle, and output the offline mapping model of the factory building environment.

[0011] More specifically, the step S102 specifically includes the following steps:

[0012] Obtain the point cloud data and operation logs collected during the operation of lidar SLAM, and extract the historical odometer and prior map information of the AGV vehicle through the operation logs;

[0013] Construct a Q-table degradation evaluation table corresponding to the original point cloud actions and original point cloud states of the logistics area according to the historical odometer and prior map information, and extract the point cloud density, distribution characteristics, plane characteristics, corner distribution and environmental change patterns of the point cloud data;

[0014] Based on the point cloud density, distribution characteristics, plane characteristics, corner distribution and environmental change patterns, select the current point cloud action and execute it to generate a new point cloud action, and continuously update and iterate the degradation discrimination state of the Q-table degradation evaluation table through the new point cloud action to obtain the maximum degradation evaluation value;

[0015] Obtain the dynamic object point cloud and noise point cloud of the AGV vehicle running in the logistics area environment, identify the dynamic object point cloud and noise point cloud to construct a degradation directed acyclic graph of the point cloud data, determine the interference degree of the dynamic object and environmental noise on the point cloud feature extraction based on the in-degree analysis of the degradation directed acyclic graph with the maximum degradation evaluation value, and perform noise suppression on the lidar SLAM for abnormal point cloud voxel density according to the interference degree.

[0016] More specifically, the obtaining the dynamic object point cloud and noise point cloud of the AGV vehicle running in the logistics area environment, identifying the dynamic object point cloud and noise point cloud to construct a degradation directed acyclic graph of the point cloud data, determining the interference degree of the dynamic object and environmental noise on the point cloud feature extraction based on the in-degree analysis of the degradation directed acyclic graph with the maximum degradation evaluation value, and performing noise suppression on the lidar SLAM for abnormal point cloud voxel density specifically includes the following steps:

[0017] Obtain the dynamic object point cloud and noise point cloud of the AGV vehicle running in the logistics scenario, identify the dynamic object point cloud and noise point cloud to obtain the dynamic object motion pattern and environmental noise characteristics, and construct a degraded directed acyclic graph of the point cloud data based on the dynamic object motion pattern and environmental noise characteristics;

[0018] Construct a statistical empty stack, calculate the in-degree of each point cloud data node in the degraded directed acyclic graph based on the maximum degradation evaluation value, traverse the in-degree of the adjacent nodes of each point cloud data node in the degraded directed acyclic graph and subtract 1. If the in-degree of a certain adjacent node changes to 0, add the node to the statistical empty stack, and finally output the in-degree statistical value. Determine the interference degree of dynamic objects and environmental noise on point cloud feature extraction according to the in-degree statistical value;

[0019] Obtain the point cloud space and spatial specifications of the point cloud data, divide the point cloud space into N point cloud voxel grids based on the spatial specification parameters, preset an abnormal density based on the interference degree, and introduce a grid density algorithm to calculate the density of each point cloud voxel in each point cloud voxel grid to obtain a number of point cloud voxel densities;

[0020] If the point cloud voxel density does not exceed the abnormal density, extract the point cloud voxel corresponding to the point cloud voxel density and define it as a representative voxel, integrate all the representative voxels, generate a filtered representative voxel, and perform noise suppression on the point cloud data obtained by lidar SLAM based on the filtered representative voxel.

[0021] More specifically, the step S104 specifically includes the following steps:

[0022] After noise suppression of the point cloud data, obtain the size, material, and glossiness of the shelf and the Manhattan orthogonal constraint of the factory building. Based on the size, material, and glossiness, perform degraded recognition of the effective point cloud and noise point cloud on the maximum degradation evaluation value to determine the point cloud generation mode of the shelf channel under the point cloud data detected by lidar;

[0023] Construct a search neighborhood of the point cloud data based on the normal direction of the point cloud of the shelf channel in the point cloud data, calculate the Manhattan neighborhood distance between each point cloud and the neighborhood point cloud in the search neighborhood according to the point cloud generation mode, and establish an accumulator with an orthogonal accumulation reference according to the Manhattan orthogonal constraint;

[0024] If the Manhattan neighborhood distance is less than the preset Manhattan neighborhood distance, add the point cloud to the accumulator for weighted calculation of Manhattan orthogonality, and finally output the most voted point cloud of the accumulator. Determine the source landmark point cloud feature of the shelf channel extracted by lidar SLAM according to the most voted point cloud.

[0025] More specifically, the step S106 specifically includes the following steps:

[0026] Obtain the real-time sensing data of the multi-modal sensor, and map the underlying state variables of the real-time sensing data regarding complementarity and redundancy to the local layer, regional layer, and global layer to construct a hierarchical factor graph;

[0027] Continuously propagate the confidence information from the bottom leaf nodes of the hierarchical factor graph to the upper local layer and the optimized confidence information from the global layer to the lower layer, and repeat the steps of the above local propagation and global propagation until the hierarchical factor graph is in an open-loop state to obtain a real-time multi-sensor fusion model;

[0028] Use the real-time multi-sensor fusion model to obtain the target landmark point cloud features of the shelf channel, calculate the point cloud data based on the dynamic parameters of the AGV vehicle to construct the optical flow constraint equation of the vehicle dynamics model, iteratively solve the optical flow constraint equation with a smoothing constraint to update the optical flow, generate the optical flow distortion amplitude, and perform calibration registration optimization on the target landmark point cloud features and the source landmark point cloud features based on the optical flow distortion amplitude for the AGV vehicle to perform positioning and navigation in the shelf channel and logistics area environment.

[0029] More specifically, the step of using the real-time multi-sensor fusion model to obtain the target landmark point cloud features of the shelf channel, calculating the point cloud data based on the dynamic parameters of the AGV vehicle to construct the optical flow constraint equation of the vehicle dynamics model, iteratively solving the optical flow constraint equation with a smoothing constraint to update the optical flow, generating the optical flow distortion amplitude, and performing calibration registration optimization on the target landmark point cloud features and the source landmark point cloud features for the AGV vehicle to perform positioning and navigation in the shelf channel and logistics area environment specifically includes the following steps:

[0030] Extract the dynamic parameters of the AGV vehicle through the historical odometer, obtain the vehicle dynamics model of the AGV vehicle based on big data, calculate the vector gradient of adjacent point clouds at different time series of the point cloud data using the dynamic parameters, and introduce an optical flow algorithm to construct the optical flow constraint equation of the vehicle dynamics model;

[0031] Iteratively solve the vector gradient with a smoothing constraint through the optical flow constraint equation to obtain the updated optical flow, and calculate the optical flow distortion amplitude between the updated optical flow and the initial optical flow;

[0032] Use the real-time multi-sensor fusion model to obtain the target landmark point cloud features of the shelf channel in a multi-modal sensing environment, find the point closest to each source landmark point cloud feature in the target point cloud feature, and calculate the rigid transformation matrix between the source landmark point cloud feature and the target landmark point cloud feature;

[0033] Optimize the rigid transformation matrix based on the optical flow distortion amplitude to obtain the corrected rigid transformation matrix. Update the position of the source landmark point cloud feature through the corrected rigid transformation matrix and re - perform the closest point registration to obtain the optimized result of the landmark point cloud feature registration in the shelf channel. Perform positioning and navigation for the AGV vehicle to operate in the shelf channel and the logistics area environment according to the optimized result of the landmark point cloud feature registration.

[0034] More specifically, the step S108 specifically includes the following steps:

[0035] Obtain the operation requirements of the AGV vehicle, construct a root node, start splitting nodes of different root branches for the point cloud data from the root node, generate an initial incremental regression tree of the sparse point cloud distribution and the environmental change law, and preset the minimum average error according to the operation requirements.

[0036] Obtain the new point cloud data when the AGV vehicle meets the operation requirements, calculate the prediction error between the new point cloud data and the point cloud data of the initial incremental regression tree, and re - split the prediction error based on the minimum average error to incrementally update the tree structure of the initial incremental regression tree, and output the offline mapping model of the plant environment.

[0037] The present invention solves the technical defects existing in the background art. The beneficial technical effects of the present invention are as follows:

[0038] Obtain the point cloud data collected by the lidar SLAM operation. Identify the Q - table degradation evaluation table through the statistical characteristics degradation of the point cloud data to obtain the maximum degradation evaluation value. Calculate the point cloud noise suppression for the lidar SLAM based on the in - degree interference degree of the maximum degradation evaluation value. After the point cloud data is noise - suppressed, construct an accumulator based on the Manhattan orthogonality constraint of the plant to perform orthogonal cumulative weighted statistics on the search neighborhood of the point cloud data to determine the source landmark point cloud feature extracted by the lidar SLAM. Obtain the real - time sensing data of the multi - modal sensor, construct a real - time multi - sensor fusion model based on the hierarchical factor graph and belief propagation calculation, use the real - time multi - sensor fusion model and the optical flow distortion amplitude of the dynamic constraint to correct and register and optimize the source landmark point cloud feature, and provide positioning and navigation for the AGV vehicle. Continuously incrementally learn and update the tree structure of the initial incremental regression tree of the point cloud data according to the operation requirements of the AGV vehicle, and output the offline mapping model of the plant environment. The present invention can accurately extract and register the point cloud for the characteristics repetition, degradation of lidar SLAM and other optical detection sensors and the prior map mismatch and random noise in the high - dynamic scene under the repeated environment of the shelf and the open logistics area for the AGV vehicle, provide the AGV vehicle with powerful environmental perception, positioning and navigation capabilities, and thus realize efficient warehousing operations and logistics task execution. Brief Description of the Drawings

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0040] Figure 1 The first method flowchart of an intelligent navigation method applicable to the shelf aisle and logistics area environment is shown;

[0041] Figure 2 The second method flowchart of an intelligent navigation method applicable to the shelf aisle and logistics area environment is shown. Detailed implementation manners

[0042] In order to more clearly understand the above objects, features, and advantages of the present invention, the present invention will be further described in detail below in conjunction with the drawings and specific implementation manners. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0043] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0044] The first aspect of the present invention provides an intelligent navigation method applicable to the shelf aisle and logistics area environment, as Figure 1 shown, including the following steps:

[0045] S102: Obtain the point cloud data collected by the lidar SLAM operation, degenerate and identify the Q-table degradation evaluation table through the statistical characteristics of the point cloud data to obtain the maximum degradation evaluation value, and calculate the point cloud noise suppression for the lidar SLAM based on the in-degree interference degree of the maximum degradation evaluation value;

[0046] S104: After the noise suppression of the point cloud data, construct an accumulator based on the Manhattan orthogonal constraint of the factory building to perform orthogonal accumulative weighted statistics on the search neighborhood of the point cloud data to determine the source landmark point cloud features extracted by the lidar SLAM for the shelf aisle;

[0047] S106: Obtain the real-time sensing data of the multi-modal sensor, construct a real-time multi-sensor fusion model based on the hierarchical factor graph and belief propagation calculation, and use the real-time multi-sensor fusion model and the optical flow distortion amplitude of the dynamic constraint to correct and register and optimize the source landmark point cloud features to provide positioning navigation for the AGV vehicle;

[0048] S108: Continuously incrementally learn and update the tree structure of the initial incremental regression tree of the point cloud data according to the operation requirements of the AGV vehicle, and output the offline mapping model of the plant environment.

[0049] More specifically, the step S102 specifically includes the following steps:

[0050] Obtain the point cloud data and operation logs collected during the operation of the lidar SLAM, and extract the historical odometer and prior map information of the AGV vehicle through the operation logs;

[0051] Construct a Q-table degradation evaluation table corresponding to the original point cloud actions and original point cloud states of the logistics area according to the historical odometer and prior map information, and extract the point cloud density, distribution characteristics, plane characteristics, corner point distribution and environmental change patterns of the point cloud data;

[0052] Based on the point cloud density, distribution characteristics, plane characteristics, corner point distribution and environmental change patterns, select and execute the current point cloud action to generate a new point cloud action, and continuously update and iterate the degradation discrimination state of the Q-table degradation evaluation table through the new point cloud action to obtain the maximum degradation evaluation value;

[0053] Obtain the dynamic object point cloud and noise point cloud during the operation of the AGV vehicle in the logistics area environment, identify the dynamic object point cloud and noise point cloud to construct a degradation directed acyclic graph of the point cloud data, analyze the in-degree of the degradation directed acyclic graph based on the maximum degradation evaluation value to determine the interference degree of the dynamic object and environmental noise on the point cloud feature extraction, and perform noise suppression on the point cloud voxel density abnormality of the lidar SLAM according to the interference degree.

[0054] It should be noted that the existing AGV vehicles rely on lidar point cloud data for environmental perception and positioning in the logistics scenario. However, when passing through an open scene, due to sparse environmental features, single geometric structure and decreased recognition of point cloud data, it is easy to cause positioning degradation problems. And the existing point cloud feature extraction methods are greatly affected by environmental factors when facing dynamic objects and environmental noise, and the point cloud data is easily interfered by noise, making it difficult to accurately distinguish dynamic objects from static environments, resulting in degradation Figure 1The reduction in consistency and the decline in repositioning accuracy lead to a significant reduction in the stability of feature extraction, affecting the navigation stability of the unmanned vehicle. Therefore, it is necessary to evaluate the degradation problem caused by the AGV vehicle passing through open scenarios during operation in the logistics scenario and suppress dynamic objects and noise when operating in a complex and changing environment. For this, this method constructs a Q-table degradation evaluation table corresponding to the original point cloud action and the original point cloud state of the logistics area based on the historical odometer of the AGV vehicle and the prior map information. The Chinese name of the Q-table degradation evaluation table is the Q-learning degradation evaluation table, which can handle the state randomness degradation evaluation of the point cloud density, distribution characteristics, plane characteristics, corner distribution, and environmental change pattern of the point cloud data. By learning the long-term cumulative reward corresponding to the Q degradation value obtained by executing a certain action in a certain state, the degradation evaluation and discrimination of the indoor structured scene can be realized, and the quality dialectics of the scene navigation degradation based on the positioning prior data can be made more accurate and reliable. Among them, the Q-table degradation evaluation table contains the states and actions of the historical odometer and the prior map information. The current point cloud action is selected through the point cloud density, distribution characteristics, plane characteristics, corner distribution, and environmental change pattern of the point cloud data. Then, the current point cloud action is executed in the Q-table degradation evaluation table to further observe the new state and reward, so that a new point cloud action can be generated to update the Q-table degradation evaluation table. Finally, the degradation discrimination state of the maximum Q value recorded in the Q-table degradation evaluation table is the best evaluation value of a scene degradation. This method can establish a degradation evaluation mechanism for the indoor logistics scene and the suppression effect of adaptive dynamic objects and noise, which helps to identify and quantify the impact of the degraded scene on the Figure 1 consistency and repositioning accuracy, improve the stability and accuracy of feature extraction, and improve the stability and robustness of the AGV vehicle in a complex environment.

[0055] More specifically, the method includes the following steps: obtaining the dynamic object point cloud and the noise point cloud of the AGV vehicle running in the logistics area environment, identifying the dynamic object point cloud and the noise point cloud to construct a directed acyclic graph of the degradation of the point cloud data, analyzing the directed acyclic graph of the degradation based on the maximum degradation evaluation value to determine the interference degree of the dynamic object and the environmental noise on the point cloud feature extraction, and performing noise suppression on the point cloud voxel density anomaly of the lidar SLAM according to the interference degree.

[0056] Obtain the dynamic object point cloud and the noise point cloud of the AGV vehicle running in the logistics scenario, identify the dynamic object point cloud and the noise point cloud to obtain the dynamic object motion pattern and the environmental noise characteristics, and construct a directed acyclic graph of the degradation of the point cloud data based on the dynamic object motion pattern and the environmental noise characteristics;

[0057] Construct a statistical empty stack, calculate the in-degree of each point cloud data node in the degraded directed acyclic graph based on the maximum degradation evaluation value, traverse the adjacent nodes of each point cloud data node in the degraded directed acyclic graph and subtract 1 from their in-degrees. If the in-degree of an adjacent node changes to 0, add this node to the statistical empty stack, and finally output the in-degree statistical value. Determine the interference degree of dynamic objects and environmental noise on point cloud feature extraction according to the in-degree statistical value;

[0058] Obtain the point cloud space and space specifications of the point cloud data, divide the point cloud space into N point cloud voxel grids based on the space specification parameters, preset the abnormal density based on the interference degree, and introduce a grid density algorithm to calculate the density of each point cloud voxel in each point cloud voxel grid, obtaining a number of point cloud voxel densities;

[0059] If the point cloud voxel density does not exceed the abnormal density, extract the point cloud voxel corresponding to this point cloud voxel density and define it as a representative voxel, integrate all representative voxels, generate a filtered representative voxel, and perform noise suppression on the point cloud data obtained by lidar SLAM based on the filtered representative voxel.

[0060] It should be noted that due to the difficulty of feature extraction caused by dynamic objects and environmental noise when the AGV vehicle runs in a complex and changeable environment in the logistics scenario, for the suppression of adaptive dynamic objects and noise, this method first identifies the dynamic object point cloud and noise point cloud of the AGV vehicle running in the logistics scenario to determine the motion mode of the dynamic object and the noise characteristics of the environment, so as to further construct a degraded directed acyclic graph of the point cloud data, revealing the degradation trend sorting of the point cloud data obtained under dynamic objects and environmental noise; then construct a statistical empty stack to store the point cloud data nodes with in-degree changes of 0, and then use the previously obtained maximum degradation evaluation value to calculate the in-degree of each point cloud data node in the degraded directed acyclic graph, so as to evaluate its sorting change under the influence of dynamic objects and environmental noise in the point cloud data degradation trend. Subtracting 1 from the in-degree of the adjacent node can achieve precursor dependence and reduce the phenomenon of mutual sorting interference of the trends between adjacent point cloud data. If the in-degree of an adjacent node changes to 0, it means that the sorting of the point cloud trend of this point cloud data node changes more severely after being affected by dynamic objects and environmental noise, and it can best reflect the interference degree of dynamic objects and environmental noise on the point cloud data, so this node is added to the statistical empty stack. Finally, all the added point cloud data nodes in the statistical empty stack are a cluster that quantifies the interference degree of dynamic objects and environmental noise on point cloud feature extraction. Through this method, the influence of dynamic objects and environmental noise on point cloud feature extraction can be effectively distinguished by using the degraded scene discrimination result and prior constraints, enabling the AGV vehicle to more quickly detect and clarify the noise degree of the extracted point cloud data in a complex and changeable logistics scenario. Thus, further stably and accurately extract environmental features.

[0061] It should be noted that after clarifying the interference degree of dynamic objects and environmental noise on the point cloud features extracted by lidar SLAM, it is necessary to suppress the motion patterns of dynamic objects and the noise characteristics of the environment. For this, this method uses the method of dividing the point cloud space of point cloud data into several point cloud voxel grids. On the one hand, it finely identifies and locates the specific positions of abnormal point clouds, making the subsequent suppression of the motion of dynamic objects and environmental noise based on the interference degree more accurate. On the other hand, it reduces the suppression of redundant points and improves the operation efficiency. Then, for the density of each point cloud voxel in each point cloud voxel grid, if the point cloud voxel density does not exceed the abnormal density preset based on the interference degree, it means that the trend sorting of the point cloud after being affected by dynamic objects and environmental noise is normal. Therefore, this point cloud can be used as a regularized sampling representative after noise removal to update the original point cloud with noise, thereby obtaining a filtered representative voxel. The point cloud data obtained by lidar SLAM is suppressed by using the filtered representative voxel, so that the suppression of dynamic objects and environmental noise is more accurate and stable. Through this method, the noise of dynamic objects and the environment can be adaptively suppressed, the robustness and accuracy of feature extraction can be improved, which helps to enhance the environmental perception ability of the unmanned vehicle, thereby improving the overall operation safety and reliability, and enabling the unmanned vehicle to perceive and navigate more reliably in a complex logistics environment.

[0062] More specifically, the step S104 specifically includes the following steps:

[0063] After noise suppression of the point cloud data, the size, material, and glossiness of the shelf and the Manhattan orthogonal constraint of the factory building are obtained. Based on the size, material, and glossiness, the degradation recognition of effective point clouds and noise point clouds is carried out on the maximum degradation evaluation value to determine the point cloud generation mode of the shelf channel under the point cloud data detected by the lidar.

[0064] Based on the normal direction of the point cloud of the shelf channel in the point cloud data, a search neighborhood of the point cloud data is constructed. According to the point cloud generation mode, the Manhattan neighborhood distance between each point cloud in the search neighborhood and the neighborhood point cloud is calculated, and an accumulator with an orthogonal accumulation reference is established according to the Manhattan orthogonal constraint.

[0065] If the Manhattan neighborhood distance is less than the preset Manhattan neighborhood distance, add this point cloud to the accumulator for weighted calculation of Manhattan orthogonality, and finally output the most voted point cloud of the accumulator. Based on the most voted point cloud, the source landmark point cloud feature of the shelf channel extracted by lidar SLAM is determined.

[0066] It should be noted that in the relocalization task in the shelf scenario, the existing technology is often interfered by the repeatability of shelf features and random noise, resulting in unstable point cloud feature extraction results and affecting the relocalization consistency. Especially in the factory environment, due to the characteristics of shelf size, material, and gloss, the lidar-detected point cloud has a non-uniform density distribution, making it difficult to accurately extract key features. In addition, traditional feature extraction methods do not fully consider the Manhattan assumption constraint in the scene, resulting in a large error in the mapping from point cloud to landmark features. In response, after suppressing the dynamic objects and environmental noise extracted from the point cloud data, this method performs degradation recognition on the degradation evaluation of the scene point cloud data based on prior information such as the size, material, and glossiness of the shelf, so as to obtain the point cloud generation mode of the shelf channel under the point cloud data detected by the lidar. Considering that there is a prior of the Manhattan assumption in the factory building, that is, the main structure lines are orthogonal to the coordinate axes, this method establishes an accumulator with an orthogonal accumulation reference based on the Manhattan orthogonal constraint of the factory building, which is used to weightedly count the parameterized point cloud data of points, lines, and planes that meet the Manhattan orthogonal constraint. If the Manhattan neighborhood distance is less than the preset Manhattan neighborhood distance, it means that the point cloud is not interfered by inconsistent relocalization under the shelf scene features repetition and random noise. Therefore, the most accurate source landmark point cloud feature of the shelf channel extracted by lidar SLAM can be further determined according to the point cloud with the most votes. By introducing the degradation scene discrimination result and combining the orthogonal constraint of the Manhattan assumption, this method can more accurately establish the mapping relationship between the point cloud and the landmark features, improve the adaptability and robustness of point cloud feature extraction, and thus enhance the relocalization accuracy and consistency.

[0067] It should be noted that the Manhattan orthogonal constraint is a geometric assumption commonly used in fields such as computer vision and robot navigation. Based on the Manhattan world assumption, that is, the main structures in the factory building (such as walls, floors, and ceilings) are usually aligned with a set of fixed and mutually orthogonal directions. In tasks such as 3D reconstruction, pose estimation, and SLAM localization and mapping in robot intelligent navigation, the Manhattan orthogonal constraint can ensure that the detected planes or edges are aligned with the main orthogonal directions, thereby improving the stability and accuracy of calculations.

[0068] Calculate the Manhattan neighborhood distance between each point cloud in the search neighborhood and the neighborhood point cloud. The corresponding distance formula is:

[0069] ;

[0070] where is the Manhattan neighborhood distance between a point cloud and a certain neighborhood point cloud, is the neighborhood distance between a point cloud and a certain neighborhood point cloud on the X-axis, is the neighborhood distance between a point cloud and a certain neighborhood point cloud on the Y-axis, is the neighborhood distance between the point cloud and a certain neighborhood point cloud on the Z-axis.

[0071] According to the Manhattan orthogonal constraint, an accumulator with an orthogonal accumulation reference is established, and the corresponding construction formula is:

[0072] ;

[0073] Among them, is the cumulative value, is the weighted weight of the point cloud for the neighborhood point cloud, is the neighborhood point cloud whose Manhattan neighborhood distance is less than the preset Manhattan neighborhood distance.

[0074] If the Manhattan neighborhood distance is less than the preset Manhattan neighborhood distance, add the point cloud to the accumulator for weighted calculation of Manhattan orthogonality. The corresponding weighted weight calculation formula is:

[0075] ;

[0076] Among them, is the weighting function, is the weight distribution adjustment factor.

[0077] Through the above formula steps, it is possible to construct an accumulator based on the Manhattan orthogonal constraint of the factory building to orthogonally accumulate and weightedly statistically search the neighborhood effect of point cloud data, so that the robot can use this Manhattan orthogonal constraint to match the line features in the image with the orthogonal directions in the world coordinate system during intelligent navigation in the factory building, so as to reduce the estimation error and improve the spatial structure accuracy and stability of point cloud feature extraction.

[0078] More specifically, the step S106 specifically includes the following steps:

[0079] Obtain the real-time sensing data of the multi-modal sensor, map the real-time sensing data about the underlying state variables of complementarity and redundancy to the local layer, regional layer, and global layer to construct a hierarchical factor graph;

[0080] Continuously propagate the confidence information from the bottom leaf nodes of the hierarchical factor graph to the upper local layer and the optimized confidence information from the global layer to the lower layer, and repeat the steps of the above local propagation and global propagation until the hierarchical factor graph is in an open-loop state to obtain a real-time multi-sensor fusion model;

[0081] The target landmark point cloud features of the shelf aisle are obtained by using a real-time multi-sensor fusion model. Based on the dynamic parameters of the AGV vehicle, the optical flow constraint equation of the vehicle dynamics model is calculated by constructing point cloud data. The optical flow constraint equation is iteratively solved with smoothing constraints to update the optical flow, and the optical flow distortion amplitude is generated. Based on the optical flow distortion amplitude, the target landmark point cloud features and the source landmark point cloud features are corrected, registered, and optimized, so as to position and navigate the AGV vehicle in the shelf aisle and the logistics area environment.

[0082] It should be noted that when the AGV vehicle registers point cloud data, there are problems such as asynchrony and random errors between multi-modal sensors in complex dynamic scenarios, resulting in inconsistent registration of point cloud data in time and space, and prone to point cloud registration errors. Secondly, the complementarity and redundancy of sensor data have not been fully exploited under different working conditions such as degradation, non-degradation, and high-speed movement, making it difficult to adapt to complex environmental changes. The existing fusion methods have insufficient modeling of noise characteristics and data delay, resulting in a decrease in system robustness. In addition, the point cloud data collected by the current AGV automated guided vehicle during driving is easily affected by changes in its own motion state, resulting in point cloud distortion. The point cloud data at different timestamps may have spatial misalignment, affecting the accuracy of environmental perception and map construction. Therefore, it is necessary to construct a real-time multi-sensor fusion model that ensures the complementarity and redundancy of multi-sensor data in scenarios such as degradation, non-degradation, and high speed, as well as a correction mechanism for dealing with point cloud data distortion under different motion states, so as to investigate the noise characteristics and data delay of different sensors in a dynamic environment and explore the correction of point cloud data under dynamic constraints.

[0083] It should be noted that for the construction of a real-time multi-sensor fusion model, this method maps the underlying state variables of the real-time sensing data of multi-modal sensors regarding complementarity and redundancy to different levels, forming a hierarchical factor graph with a hierarchical structure, so that local inference and global optimization between the data of different sensors can be carried out separately, reducing asynchronous and random errors. Then, local belief propagation is performed, that is, the confidence information is locally propagated from the bottom leaf nodes of the hierarchical factor graph to the upper layer, and local information fusion is carried out between the bottom sensors to improve the reliability of local state estimation. Then, the optimized confidence information is propagated from the global layer to the lower layer, so that the local estimation is adjusted under global constraints, realizing global consistency optimization, improving the overall fusion synchronization consistency of different sensor data, and avoiding the generation of local random errors during point cloud registration. If the factor graph has a closed-loop structure, it means that the data fusion cost of multi-sensors is large and it is difficult to ensure the accuracy of large-scale data fusion. Therefore, the factor graph is in an open-loop state during multiple rounds of iterative optimization, and finally a real-time multi-sensor fusion model based on the hierarchical factor graph and belief propagation can be generated. Through this method, a real-time multi-sensor data fusion mechanism based on the factor graph and belief propagation model can be established, making full use of the information complementarity between multi-modal sensors, improving the accuracy and stability of sensing data fusion in a dynamic environment, and eliminating asynchronous and random errors between multi-modal sensors in a complex dynamic scenario.

[0084] More specifically, the method uses a real-time multi-sensor fusion model to obtain the target landmark point cloud features of the shelf aisle, calculates the point cloud data based on the dynamic parameters of the AGV vehicle to construct an optical flow constraint equation of the vehicle dynamics model, smoothly constrains and iteratively solves the optical flow constraint equation to update the optical flow, generates an optical flow distortion amplitude, and corrects and registers and optimizes the target landmark point cloud features and the source landmark point cloud features based on the optical flow distortion amplitude, so as to position and navigate the AGV vehicle running in the shelf aisle and the logistics area environment, as Figure 2 shown, and specifically includes the following steps:

[0085] S202: Extract the dynamic parameters of the AGV vehicle through the historical odometer, obtain the vehicle dynamics model of the AGV vehicle based on big data, calculate the vector gradient of adjacent point clouds at different time series of the point cloud data using the dynamic parameters, and introduce an optical flow algorithm to construct an optical flow constraint equation of the vehicle dynamics model;

[0086] S204: Iteratively solve the vector gradient through the optical flow constraint equation with smooth constraints to obtain the updated optical flow, and calculate the optical flow distortion amplitude between the updated optical flow and the initial optical flow;

[0087] S206: Obtain the target landmark point cloud features in the multi-modal sensing environment of the shelf aisle using the real-time multi-sensor fusion model, find the points closest to each source landmark point cloud feature in the target point cloud features, and calculate the rigid transformation matrix between the source landmark point cloud features and the target landmark point cloud features;

[0088] S208: Optimize the rigid transformation matrix based on the optical flow distortion amplitude correction to obtain the corrected rigid transformation matrix. Update the positions of the source landmark point cloud features through the corrected rigid transformation matrix and perform nearest point registration again to obtain the optimized result of the landmark point cloud feature registration of the shelf aisle. Position and navigate the AGV vehicle in the shelf aisle and logistics area environment according to the optimized result of the landmark point cloud feature registration.

[0089] It should be noted that for the problem of point cloud data distortion caused by motion changes during the driving of the AGV, this method extracts the dynamic parameters of the AGV vehicle based on the historical odometer and calculates the vector gradient of adjacent point clouds at different time sequences of the point cloud data. The vector gradient includes the spatial gradient and time gradient of the point cloud data, and is used to capture the dynamic information of the point cloud data changing with time. Then, an optical flow constraint equation of the vehicle dynamics model is constructed based on the vector gradient. Thus, the distortion mode of the point cloud data under different motion states can be analyzed. Smoothly constraining and iteratively solving this optical flow constraint equation is to ensure that the optical flow is consistent with the gray-scale change of the point cloud, which helps to capture the real motion trajectory of the AGV vehicle driving in the image. The optical flow estimation after each iteration optimization becomes more accurate and can express the smoothness of the point cloud data. Based on the updated optical flow, the distortion amplitude of the point cloud data can be further revealed. Finally, this optical flow distortion amplitude is the specific quantification of the point cloud data distortion and is also an important basis for correction. Through this method, it is possible to explore point cloud correction under dynamic constraints, which helps to improve the spatial consistency of the point cloud data, can effectively compensate for errors caused by motion, and enhance the perception accuracy of the AGV vehicle in complex environments, providing reliable data support for high-precision map construction and autonomous navigation. In vehicle dynamics, the optical flow is not only affected by the point cloud gradient but also related to the motion of the AGV vehicle itself. According to rigid body kinematics, the optical flow motion can be derived from the vehicle's dynamics model, and the corresponding optical flow expression formula is:

[0090] ;

[0091] where u and v are the optical flow components, Z is the depth from the coordinate center point to the lidar, f is the sensing distance of the lidar, ω is the angular velocity of the AGV vehicle, X and Y are the relative positions of the scene currently observed by the vehicle, V is the translational velocity component of the vehicle in the world coordinate system, and x and y are the directional velocities.

[0092] Substitute the optical flow expression of the vehicle dynamics model into the derivation of the optical flow algorithm, and the corresponding optical flow equation derivation formula is:

[0093] ;

[0094] Among them, is the time gradient.

[0095] During the point cloud acquisition process, due to the movement of the sensor under the vehicle dynamics conditions followed by the AGV vehicle, the point clouds sampled at different times will be spatially inconsistent, resulting in distortion. The core reason why the optical flow algorithm can be used for point cloud distortion correction lies in its motion estimation ability. Especially when using a SLAM lidar or an RGB-D camera, it can calculate the motion vectors of each point in the point cloud of the acquired image and estimate the true position of the spatial point cloud, thereby compensating for the errors caused by motion. The optical flow algorithm is not limited to the optical flow estimation of image pixels. It can also be used for the global point cloud motion correction of the rigid body motion of an autonomous vehicle and for capturing the local non-rigid correction of the point cloud in complex motions such as acceleration and rotation. Therefore, the optical flow algorithm can estimate the non-uniform distribution of the point cloud under the vehicle dynamics constraints, calculate the point cloud distortion, and can combine with IMU, SLAM or deep learning models to improve the distortion correction accuracy, making the point cloud distortion correction more consistent and accurate.

[0096] It should be noted that in practical applications, the existing point cloud feature registration technology is often affected by environmental dynamic changes and sensor noise, resulting in registration errors. These errors mainly come from sensor errors and environmental changes. Especially in complex or dynamic environments, traditional registration methods are difficult to cope with the uncertainties brought by the environment and sensors. Therefore, the existing registration algorithms have poor adaptability and robustness in dynamic environments, are prone to large misregistration errors, and affect the accuracy and reliability of point cloud data. In response to this, this method obtains the target landmark point cloud features in the multi-modal sensing environment of the shelf channel by using a real-time multi-sensor fusion model, and then finds the points closest to each source landmark point cloud feature in the target point cloud features, making the source point cloud more tend to the points in the target point cloud to establish a one-to-one correspondence relationship, minimizing the error of the final point cloud data registration result. Then, calculate the rigid transformation matrix between the source landmark point cloud features and the target landmark point cloud features. Since the point cloud data was distorted before, this method corrects the rigid transformation matrix by using the calculated optical flow distortion amplitude, providing more accurate rotation and translation transformations for registration. Especially in non-rigid or noisy point cloud data, more reliable registration results can be obtained. Finally, by updating the positions of the source landmark point cloud features with the corrected rigid transformation matrix and re-performing the closest point registration, more accurate positioning and navigation can be provided for the AGV vehicle to operate in the shelf channel and logistics area environment. Through this method, multiple information sources can be integrated, effectively reducing the propagation and amplification of errors, thereby improving the accuracy and stability of point cloud registration, enabling the optimization strategy of dynamic constraints to better handle the uncertainties brought by environmental changes and sensor noise in practical applications, and enhancing the robustness of the registration algorithm. Among them, the rigid transformation matrix is a matrix used to describe the position and orientation transformation of an object in space. The rigid transformation preserves the shape and size of the object and is usually used to represent translation and rotation operations without involving scaling or deformation.

[0097] Before the rigid body transformation, it is first necessary to calculate the centroid between the source landmark point cloud features and the target landmark point cloud features. The corresponding formula is:

[0098] ;

[0099] Among them, is the centroid of the source landmark point cloud, is the centroid of the target landmark point cloud, n is the total number of points in the point cloud, , are the point clouds.

[0100] Then, the covariance matrix between the source landmark point cloud features and the target landmark point cloud features. The corresponding formula is:

[0101] ;

[0102] Among them, H is the covariance matrix, , is a column vector, is a row vector, and T is a rigid transformation matrix.

[0103] Finally, the rigid transformation matrix can be generated according to the rotation and translation of the covariance matrix. The corresponding rigid transformation matrix formula is:

[0104] ;

[0105] Among them, R is the rotation matrix and t is the translation vector. The rigid transformation matrix is a 4×4 matrix, which contains the rotation matrix R and the translation vector t.

[0106] More specifically, the step S108 specifically includes the following steps:

[0107] Obtain the operation requirements of the AGV vehicle, construct the root node, start from the root node to split the point cloud data into nodes of different root branches, generate the initial incremental regression tree of the sparse point cloud distribution and the environmental change law, and preset the minimum average error according to the operation requirements;

[0108] Obtain the new point cloud data when the AGV vehicle reaches the operation requirements, calculate the prediction error between the new point cloud data and the point cloud data of the initial incremental regression tree, and re-split the prediction error based on the minimum average error to incrementally update the tree structure of the initial incremental regression tree, and output the offline mapping model of the plant environment.

[0109] It should be noted that when dealing with the object distribution and occlusion problems in the shelf environment, especially in the environment where AGV vehicles operate for a long time, the sparsity of point cloud data will affect the accuracy of environmental perception and map construction. It is difficult to effectively cope with the dynamic changes and occlusion problems in the environment, resulting in untimely map updates and possibly affecting the repositioning accuracy. In addition, traditional point cloud map construction methods lack adaptability to long-term environmental changes and are difficult to meet the high requirements for accuracy and consistency in the actual application of AGV vehicles. Therefore, this method analyzes the distribution characteristics of sparse point cloud data and the environmental change rules by splitting the point cloud data according to the operation demand nodes of the AGV vehicle, so as to generate an initial incremental regression tree of the sparse point cloud distribution and environmental change rules. Each node split can divide the point cloud data into more representative subsets, enabling each leaf node to more accurately reflect the distribution characteristics of the sparse point cloud data and the environmental change rules, and ensuring the construction accuracy of the offline mapping model. Subsequently, the prediction error between the newly generated point cloud data and the point cloud data of the initial incremental regression tree is calculated based on the newly generated point cloud data that meets the operation demand. As new data arrives, the existing model is updated and adjusted by minimizing the average error increment that meets the operation demand, rather than retraining the entire tree. This can ensure that the offline mapping of the factory environment maximally meets the operation needs of the AGV vehicle, optimize the sparse point cloud phenomenon caused by the repeated distribution and occlusion of objects in the shelf environment of the AGV vehicle, and make the finally output offline mapping model of the factory environment more stable and accurate. Through this method, the consistency and repositioning accuracy of the factory offline map can be effectively improved, and rapid and accurate map updates can also be achieved in a complex dynamic environment, ensuring that the AGV vehicle has stable and reliable navigation capabilities during long-term operation.

[0110] In addition, the intelligent navigation method applicable to the shelf channel and logistics area environment further includes the following steps:

[0111] Obtain the offline mapping model of the factory environment when the AGV vehicle operates for a long time on different time scales, continuously calculate the deviation of the offline mapping model of the factory environment based on different time scales to obtain the cumulative model deviation, and use the cumulative model deviation to update the historical minimum cumulative deviation of the previous time scale to generate a detection statistic;

[0112] If the detection statistic is greater than the preset detection statistic, recalculate the mean value or reset the statistic. If it is less than, continue to accumulate and calculate until the online long-term repositioning result of the sparse point cloud data is finally obtained;

[0113] Introduce the principal component analysis algorithm to calculate the normal direction of the point cloud data, calculate the included angle value between the normal directions of each point cloud and its adjacent point cloud to obtain the normal included angle value, and preset the normal included angle threshold according to the online long-term repositioning result;

[0114] If the value of the normal angle is greater than the normal angle threshold, then flip the point cloud normal of this normal angle value, and repeat the above steps of flipping the normal angle of adjacent point clouds until new normals are generated for all point clouds, obtaining a local rejection point cloud group.

[0115] It should be noted that in the application of sparse point cloud maps, especially during long-term operation, map mismatch and drift phenomena often occur as the scene changes. Due to environmental changes, sensor errors, and the uncertainty of data collection, it is difficult to effectively cope with these changes, resulting in a gradual decline in the accuracy and stability of the map, making it difficult to detect and correct these drift problems in real time, and affecting the long-term reliability and self-adaptability of the system. In addition, due to the interference of dynamic objects, the accuracy and stability of point cloud data are affected. This instability leads to incorrect matching of point cloud features, thus affecting the subsequent mapping and positioning accuracy. In response to this, this method calculates the model deviation of the offline mapping model of the factory environment during the long-term operation of the AGV vehicle at different time scales by cumulative calculation, and identifies whether there are abnormal fluctuations or changes in long-term repositioning in the point cloud data; and uses this cumulative model deviation to update the historical minimum cumulative deviation of the previous time scale, ensuring that each time the detection statistic is calculated, it is compared with the historical lowest point, which helps to distinguish normal fluctuations from abnormal fluctuations; thereby detecting the change patterns of sparse point cloud data at different time scales. If the detection statistic is greater than the preset detection statistic, it indicates that there is a long-term repositioning anomaly in the point cloud data, so the mean value needs to be recalculated or the statistic needs to be reset. Otherwise, continue the cumulative calculation to achieve the elimination of anomalies in online long-term repositioning. Then, this method calculates the normal of the point cloud data and calculates the normal angle value between each point cloud and its adjacent point cloud. If the normal angle value is greater than the preset normal angle threshold of the online long-term repositioning result, it indicates that there is a mismatch phenomenon in the sparse point cloud data of this normal angle value in the dynamic environment. Therefore, the point cloud normal of this normal angle value can be further flipped and the above steps of flipping the normal angle of adjacent point clouds are repeated until new normals are generated for all point clouds to obtain the rejection point cloud group of local mismatch in the point cloud data. Through this method, the output result of the offline mapping model can be used to introduce an online environmental change monitoring mechanism, explore the long-term stability of point cloud features under time consistency constraints, significantly improve the stability, accuracy, and robustness of the sparse point cloud map during long-term operation, thus solving the mismatch and drift problems existing in the prior art; at the same time, designing a local point cloud rejection algorithm to exclude the input effect of mismatched features and exclude unreliable point cloud data input caused by dynamic objects or mismatched features, effectively improving the stability of point cloud feature extraction and matching.

[0116] The above are only the specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An intelligent navigation method suitable for shelf aisles and logistics area environments, characterized in that: The following steps are involved: S102: Obtain point cloud data collected by the laser radar SLAM operation, identify the Q-table degradation evaluation table through the statistical characteristic degradation of the point cloud data, obtain the maximum degradation evaluation value, and perform point cloud noise suppression on the laser radar SLAM based on the in-degree interference degree calculation of the maximum degradation evaluation value; The Q-table degradation evaluation table includes the status and action of the historical odometer and prior map information. The Q-learning degradation evaluation table corresponding to the original point cloud action and the original point cloud state of the logistics area is constructed according to the historical odometer and prior map information of the AGV vehicle. The degradation evaluation identification of indoor structured scenes is realized by learning the long-term cumulative reward corresponding to the Q degradation value that can be obtained by performing a certain action in a certain state. The in-degree interference degree is specifically: using the previously obtained maximum degradation evaluation value to calculate the in-degree of each point cloud data node in the degraded directed acyclic graph, so as to evaluate the ranking change under the influence of dynamic objects and environmental noise in the degradation trend of the point cloud data; S104: After noise suppression of the point cloud data, an accumulator is constructed based on the Manhattan orthogonal constraint of the factory to perform orthogonal accumulation weighted statistics on the search neighborhood of the point cloud data to determine the source landmark point cloud features of the shelf channel extracted by the LiDAR SLAM; The Manhattan orthogonal constraint is specifically: based on the Manhattan world assumption, the main characteristic structures in the factory are aligned with a set of fixed, mutually orthogonal directions; the main characteristic structures include walls, floors and ceilings; S106: Acquire real-time sensing data from multimodal sensors, build a real-time multi-sensor fusion model based on hierarchical factor graphs and belief propagation calculations, and use the real-time multi-sensor fusion model and the optical flow distortion amplitude of dynamic constraints to correct and optimize the source landmark point cloud features, providing positioning and navigation for AGV; The construction process of the hierarchical factor graph is as follows: mapping the underlying state variables of the real-time sensor data regarding complementarity and redundancy to a local layer, a regional layer, and a global layer; S108: Continuously incrementally learn and update the tree structure of the initial incremental regression tree of the point cloud data according to the operation requirements of the AGV vehicle, and output an offline mapping model of the factory environment.

2. The intelligent navigation method applicable to shelf aisles and logistics area environments according to claim 1, characterized in that: The step S102 specifically includes the following steps: Obtain the point cloud data and operation log collected during the operation of the LiDAR SLAM, and extract the historical odometer and prior map information of the AGV vehicle through the operation log; According to the historical odometer and prior map information, a Q-table degradation evaluation table corresponding to the original point cloud action and the original point cloud state of the logistics area is constructed, and the point cloud density, distribution characteristics, plane characteristics, corner point distribution and environmental change pattern of the point cloud data are extracted; Based on the point cloud density, distribution characteristics, plane characteristics, corner point distribution and environment change mode, the current point cloud action is selected and executed to generate a new point cloud action, and the degradation identification state of the Q-table degradation evaluation table is continuously updated and iterated through the new point cloud action to obtain the maximum degradation evaluation value; The dynamic object point cloud and noise point cloud of the AGV vehicle running in the logistics area environment are obtained, the dynamic object point cloud and the noise point cloud are identified to construct a degraded directed acyclic graph of the point cloud data, and the interference degree of dynamic objects and environmental noise on the point cloud feature extraction is determined based on the in-degree analysis of the maximum degradation assessment value, and the noise of the abnormal point cloud voxel density is suppressed for the lidar SLAM according to the interference degree.

3. The intelligent navigation method applicable to shelf aisles and logistics area environments according to claim 2, characterized in that: The method obtains the dynamic object point cloud and the noise point cloud of the AGV vehicle running in the logistics area environment, identifies the dynamic object point cloud and the noise point cloud to construct a degraded directed acyclic graph of the point cloud data, analyzes the degraded directed acyclic graph based on the maximum degradation assessment value in-degree to determine the interference degree of the dynamic object and the environmental noise on the point cloud feature extraction, and performs noise suppression of the abnormal point cloud voxel density on the laser radar SLAM according to the interference degree, specifically including the following steps: Obtain the dynamic object point cloud and noise point cloud of the AGV vehicle running in the logistics scene, identify the dynamic object point cloud and the noise point cloud to obtain the dynamic object motion mode and environmental noise characteristics, and construct a degenerate directed acyclic graph of the point cloud data based on the dynamic object motion mode and environmental noise characteristics; Construct a statistical empty stack, calculate the in-degree of each point cloud data node in the degraded directed acyclic graph based on the maximum degradation evaluation value, traverse the in-degree of the adjacent nodes of each point cloud data node in the degraded directed acyclic graph and reduce it by 1. If the in-degree of an adjacent node changes to 0, add the node to the statistical empty stack, and finally output the in-degree statistics. According to the in-degree statistics, determine the interference degree of dynamic objects and environmental noise on point cloud feature extraction; Acquire the point cloud space and space specifications of the point cloud data, divide the point cloud space into N point cloud voxel grids based on the space specification parameters, preset the abnormal density based on the interference degree, introduce the grid density algorithm to calculate the density of each point cloud voxel in each point cloud voxel grid, and obtain a number of point cloud voxel densities; If the point cloud voxel density does not exceed the abnormal density, the point cloud voxel corresponding to the point cloud voxel density is extracted and defined as the representative voxel. All representative voxels are integrated to generate filtered representative voxels. Noise suppression is performed on the point cloud data obtained by the lidar SLAM based on the filtered representative voxels.

4. The intelligent navigation method applicable to shelf aisles and logistics area environments according to claim 1, characterized in that: The step S104 specifically includes the following steps: After noise suppression of the point cloud data, the size, material and glossiness of the shelf and the Manhattan orthogonal constraints of the factory are obtained, and the degradation of the effective point cloud and the noise point cloud is identified based on the maximum degradation evaluation value of the size, material and glossiness to determine the point cloud generation mode of the shelf channel under the point cloud data obtained by laser radar detection; A search neighborhood of point cloud data is constructed based on the normal direction of the point cloud of the shelf channel in the point cloud data. The Manhattan neighborhood distance between each point cloud in the search neighborhood and the neighboring point cloud is calculated according to the point cloud generation mode, and an accumulator with an orthogonal accumulation benchmark is established according to the Manhattan orthogonal constraint. If the Manhattan neighborhood distance is less than the preset Manhattan neighborhood distance, the point cloud is added to the accumulator for weighted calculation of Manhattan orthogonality, and finally the accumulator's most voting point cloud is output. The source landmark point cloud features of the shelf channel extracted by the lidar SLAM are determined based on the most voting point cloud.

5. The intelligent navigation method applicable to shelf aisles and logistics area environments according to claim 1, characterized in that: The step S106 specifically includes the following steps: Acquire real-time sensing data from multimodal sensors; Continuously propagate confidence information from the bottom leaf nodes of the hierarchical factor graph to the upper local layer and propagate the optimized confidence information from the global layer to the lower layer, repeat the above steps of local propagation and global propagation until the hierarchical factor graph is in an open-loop state, and obtain a real-time multi-sensor fusion model; A real-time multi-sensor fusion model is used to obtain the target landmark point cloud features of the shelf aisle, and the optical flow constraint equation of the vehicle dynamics model is constructed based on the point cloud data calculated based on the dynamic parameters of the AGV vehicle. The optical flow constraint equation is solved by smooth constraint iteration to update the optical flow and generate the optical flow distortion amplitude. Based on the optical flow distortion amplitude, the target landmark point cloud features and the source landmark point cloud features are corrected, registered and optimized to perform positioning and navigation for the AGV vehicle running in the shelf aisle and logistics area environment.

6. The intelligent navigation method applicable to shelf aisles and logistics area environments according to claim 5, characterized in that: The method uses a real-time multi-sensor fusion model to obtain the target landmark point cloud features of the shelf channel, calculates the point cloud data based on the dynamic parameters of the AGV vehicle to construct the optical flow constraint equation of the vehicle dynamics model, solves the optical flow constraint equation by smooth constraint iteration to update the optical flow, generates the optical flow distortion amplitude, and corrects and aligns the target landmark point cloud features with the source landmark point cloud features based on the optical flow distortion amplitude, so as to locate and navigate the AGV vehicle in the shelf channel and logistics area environment, specifically including the following steps: The dynamic parameters of the AGV vehicle are extracted through the historical odometer, and the vehicle dynamics model of the AGV vehicle is obtained based on big data. The vector gradients of adjacent point clouds of the point cloud data at different time sequences are calculated using the dynamic parameters, and the optical flow algorithm is introduced to construct the optical flow constraint equation of the vehicle dynamics model; Iteratively solving the smooth constraint of the vector gradient through the optical flow constraint equation to obtain an updated optical flow, and calculating the optical flow distortion amplitude between the updated optical flow and the initial optical flow; Use the real-time multi-sensor fusion model to obtain the target landmark point cloud features in the shelf channel in a multi-modal sensing environment, find the point closest to the target point cloud features and each source landmark point cloud feature in the point cloud data, and calculate the rigid transformation matrix between the source landmark point cloud features and the target landmark point cloud features; The rigid transformation matrix is ​​optimized based on the optical flow distortion amplitude correction to obtain a corrected rigid transformation matrix. The position of the source landmark point cloud feature is updated through the corrected rigid transformation matrix and the nearest point registration is re-performed to obtain the landmark point cloud feature registration optimization result of the shelf channel. According to the landmark point cloud feature registration optimization result, positioning and navigation are performed for AGV vehicles running in the shelf channel and logistics area environment.

7. The intelligent navigation method applicable to shelf aisles and logistics area environments according to claim 1, characterized in that: The step S108 specifically includes the following steps: Obtain the operation requirements of the AGV vehicle, build the root node, split the point cloud data into nodes with different root branches starting from the root node, generate the initial incremental regression tree of the sparse point cloud distribution and the law of environmental changes, and preset the minimum average error according to the operation requirements; Obtain new point cloud data that meets the operation requirements of the AGV vehicle, calculate the prediction error between the new point cloud data and the point cloud data of the initial incremental regression tree, re-split the prediction error based on minimizing the average error, update the tree structure of the initial incremental regression tree incrementally, and output the offline mapping model of the factory environment.

Images