AGV outdoor positioning switching method, computer device and program product

Abstract

The present invention discloses an AGV outdoor positioning switching method, a computer device and a program product, which construct a point cloud map of the outdoor work scene and calibrate the area where the GPS signal is unreliable; judge the availability of the GPS positioning information at the initial moment, select the corresponding positioning initialization method according to the judgment result of the availability of the GPS positioning information at the initial moment and complete the vehicle positioning initialization; during the vehicle's travel, the availability of the GPS positioning information and the laser radar positioning information at the current moment is continuously judged, and the positioning mode switching is decided in combination with the currently selected positioning mode. Aiming at the situation where the GPS signal is unreliable in the outdoor navigation scene of the AGV, the present invention provides a positioning switching method and system based on GPS and laser radar, which can ensure the robustness of the positioning signal in the outdoor scene and guarantee the continuity and stability of the execution of the navigation task.

Description

Technical Field

[0001] The present invention relates to the field of positioning technology, and in particular to an AGV outdoor positioning switching method, a computer device and a program product. Background Art

[0002] The autonomous navigation operations of AGV and various mobile robots play an important role in social production and life, and are widely used in various scenarios in industry, agriculture, and services. The working scenarios of AGV and various robots are not limited to indoor scenes such as factories, workshops, and shopping malls. For example, AGVs are gradually being used in large farms to transport crops.

[0003] Although outdoor positioning can be solved by the positioning information provided by GPS most of the time, the GPS signal will be affected by weather and various obstructions. The simple outdoor positioning method based on GPS can no longer meet people's needs for the continuity and robustness of navigation tasks. Therefore, under the premise of GPS positioning information, adding other sensors to assist positioning has become a research hotspot, including auxiliary positioning methods based on wheel odometers, IMUs, and cameras. After investigation, it was found that the auxiliary positioning method based on wheel odometers and IMUs can provide the posture information required for positioning when there is no GPS positioning information, but its positioning error gradually accumulates over time and eventually becomes unreliable. In addition, the wheel odometer is prone to slip on uneven roads, which has a great impact on positioning accuracy. The existing visual positioning solution cannot guarantee the positioning accuracy required for navigation. Therefore, LiDAR is selected as an auxiliary positioning method, and posture estimation is performed through point cloud matching to ensure the supply of positioning information when GPS signals are unreliable.

[0004] The laser odometry matches the point cloud data scanned by the current frame with the point cloud data of the previous frame, finds the correspondence between the two frames of point cloud data, and solves the optimal posture change relationship by minimizing the distance between matching points, thereby achieving real-time posture estimation. However, since there is a certain probability of mismatching in inter-frame matching, the pose estimation error provided accumulates seriously over time. As a result, an optimized laser odometry based on inter-frame matching and local sub-graph matching is produced. The probability of mismatching is reduced by matching with the sub-graph constructed in real time, which provides the accuracy of the laser odometry, but there is still the problem of error accumulation. Therefore, designing a matching method to overcome the problem of pose estimation error accumulation has become the key to solving this engineering problem.

[0005] The existing positioning switching method is essentially a fusion positioning method based on Kalman filtering. When a certain positioning information source is unreliable, it will be eliminated. In the fusion positioning, the GPS positioning data is of high precision, while the positioning accuracy of the real-time mapping positioning based on lidar is low. When the GPS positioning signal is unreliable, the GPS data will be eliminated or the sampling frequency of the GPS data will be reduced in the fusion positioning based on Kalman filtering. At this time, the SLAM positioning based on lidar is the main positioning source, but SLAM is essentially a posture estimation method. Even if there is loop detection for posture correction, it cannot eliminate the mapping and positioning errors caused by error accumulation. If the GPS signal is unreliable for a long time, the positioning accuracy will inevitably gradually decrease or even become unavailable. The present invention is based on matching the lidar point cloud data with the prior truth map (constructed by the GPS true posture), which is an absolute positioning without error accumulation. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide an AGV outdoor positioning switching method, a computer device and a program product to continuously and stably provide reliable positioning information for AGV to perform outdoor navigation tasks in view of the shortcomings of the existing technology.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: an AGV outdoor positioning switching method, comprising the following steps:

[0008] S1. Collect AGV working scene data and establish a priori global point cloud map of the working scene;

[0009] S2. constructing a representation method of an unreliable GPS signal area using the prior global point cloud map of the working scene to determine the availability status of the current GPS signal;

[0010] S3. If the current GPS signal is unavailable, the availability of the lidar positioning information is determined. If the lidar positioning information is available, switch to lidar positioning. If the lidar is used to complete the positioning initialization, when the GPS signal is available and the difference between the lidar posture and the GPS posture is less than a given threshold, switch to GPS positioning.

[0011] The present invention provides feature information for a laser radar-based positioning solution when GPS signals are unavailable by mapping the working scene and storing key frame point cloud data and key frame pose information; provides auxiliary information for determining GPS availability during navigation operations by calibrating GPS signal blocking areas in the constructed global point cloud map; provides strong guarantees for accurate determination of GPS signal availability and validity of navigation positioning information by constructing criteria for GPS signal availability; ensures that AGV can complete positioning initialization under any conditions through the design of two positioning initialization solutions; and proposes a positioning mode selection and switching mechanism during AGV navigation operations based on the availability determination results of GPS signals and laser radar-based positioning information, thereby ensuring smooth and stable execution of navigation operation tasks.

[0012] The specific implementation process of step S1 includes:

[0013] Control the AGV to move forward at a speed less than the first set threshold, and record the timestamp t of the AGV entering the blocked area s And the timestamp t of the AGV leaving the blocked area e ;

[0014] Different mapping pose acquisition methods are selected according to the timestamp of the marked signal-blocked area. If the current data is not in the signal-blocked area, the pose information provided by GPS is used for mapping; if the current data belongs to the signal-blocked area, the pose information provided by point cloud matching is used for mapping.

[0015] Determine the moving distance of the vehicle according to the posture information obtained in real time, and save the key frame posture and key frame point cloud when the moving distance is greater than a second set threshold;

[0016] Splice the key frame poses and key frame point clouds to obtain a global point cloud map.

[0017] The present invention marks the data of the GPS signal blocked area, selects the pose provided by point cloud registration for mapping in the signal blocked area, and selects the pose based on GPS for mapping in the non-blocked area. In this way, the pose information source is flexibly selected during mapping, which improves the high quality of mapping. For matching positioning based on a priori map, the accuracy of the map largely determines the accuracy of positioning. Therefore, this mapping method based on different pose sources provides a strong guarantee for later map-based positioning.

[0018] In step S2, the specific implementation process of constructing a representation method of the GPS signal unreliable area using the prior global point cloud map of the working scene includes:

[0019] Read the global point cloud map, calibrate the m signal blocking areas in the global point cloud map, mark the boundary points of the signal blocking areas, and record the coordinate information of the boundary points in the map;

[0020] Dilate the signal occlusion area based on the boundary points;

[0021] For any signal blocking area, suppose n boundary points are marked, and 3 points are randomly selected from the n boundary points to form a set of arrangements. The radius of the circumscribed circle corresponding to each arrangement is calculated, and the arrangement with the largest circumscribed circle radius is selected as the marking boundary point of the signal blocking area; n>3;

[0022] Calculate the coordinates (x block1 ,y block1 ) and the outer circle radius r1;

[0023] The above expansion operation is performed on all signal occlusion areas in the global point cloud map to obtain the center coordinate sequence of the expansion area (x block1 ,y block1 ),(x block2 ,y block2 ),...,(x blockm ,y blockm ) and the expansion zone radius sequence r1,r2,...,r m ,Each expansion center coordinate plus the expansion area radius corresponding to the expansion center represents a circular GPS signal unreliable area in the prior map.

[0024] The present invention performs an expansion operation on an unreliable area in a map, thereby preventing a positioning signal from being interfered with when a vehicle is close to an unreliable GPS signal area, thereby ensuring the stability and robustness of the positioning signal.

[0025] In step S2, the specific implementation process of determining the available status of the current GPS signal includes:

[0026] Read the GPS raw data, parse the GPS raw data, and obtain the current positioning status flag information. The flag information is true when the vehicle can obtain its own real-time position and posture. At this time, the GPS reliability flag A=1, otherwise the GPS reliability flag A=0;

[0027] According to the positional relationship between the current posture and the calibrated GPS signal unreliable area, the GPS reliability flag B is obtained;

[0028] The GPS signal reliability flag C is obtained based on the current AGV position information in the map, vehicle speed information and data jump threshold;

[0029] If the GPS reliability determination flags A=1, B=1, and C=1 are satisfied at the same time, the GPS availability flag is set to true, that is, the GPS signal is available; otherwise, it is set to false, that is, the GPS signal is unavailable.

[0030] The present invention introduces a variety of GPS availability judgment criteria. The judgment process does not simply rely on the positioning status flag information provided by the GPS itself, but also introduces judgment criteria based on unreliable areas of prior map signals and judgment criteria based on the continuity of vehicle driving posture, so that the judgment and decision-making process of positioning switching is more confident and more robust, and will not make wrong switching decisions due to misjudgment of a single criterion.

[0031] According to the positional relationship between the current posture and the calibrated GPS signal unreliable area, the specific implementation process of obtaining the GPS reliability flag B includes:

[0032] 1) Read the current position information (pose.x, pose.y) of the AGV in the global point cloud map, the coordinate sequence of the signal occlusion expansion center, and the expansion area radius sequence r i , where i = 1, 2, 3, ..., m;

[0033] 2) Calculate the distance d between the current position of the AGV and the expansion center of the signal occlusion area i in the expansion center coordinate sequence i ;

[0034] 3) If d i <r i , then the AGV is in the signal unreliable area, the reliability flag B = 0, and the calculation of d is stopped i If d i >r i And the signal occlusion expansion center coordinate sequence traversal is completed, then the reliability flag B = 1; if d i >r i However, if the signal occlusion expansion center coordinate sequence is not traversed, then return to step 2).

[0035] The present invention calibrates the GPS signal unreliable area in the priori map, and can provide a criterion based on the priori map for the availability judgment of the GPS signal during the positioning switching process, so that the positioning switching judgment is more complete.

[0036] The specific implementation process of obtaining the GPS signal reliability flag C based on the current AGV position information in the map, vehicle speed information and data jump threshold includes:

[0037] Calculate the current linear velocity v and steering angular velocity w of the AGV;

[0038] Read the current AGV's position information in the map (pose.x, pose.y, heading) and the previous AGV's position information in the map (pose.x last ,pose.y last ,heading last );

[0039] Calculate the position change threshold H d :H d = k × v × T; where T is the time interval for GPS signal release, and k is the threshold expansion coefficient;

[0040] Calculate the yaw angle change threshold H h :H h = k × w × T;

[0041] Calculate the position difference d and posture difference h of the AGV between the current moment and the previous moment:

[0042]

[0043] h=heading-heading last ;

[0044] If d <H d And h <H h , then the GPS signal reliability flag C is set to 1, otherwise the GPS signal reliability flag C is set to 0; if the GPS reliability judgment flags A=1, B=1, and C=1 are satisfied at the same time, then the GPS availability flag is set to true, otherwise it is set to false.

[0045] The present invention analyzes the continuity of the vehicle's posture during driving, can monitor the abnormal jump of the vehicle's posture, infer the occurrence of abnormal current positioning information, and thus perform corresponding positioning switching operations. The introduction of this criterion adds kinematic constraints to the positioning switching judgment, making the positioning switching judgment more complete.

[0046] In step S3, the specific implementation process of using laser radar to complete positioning initialization includes:

[0047] Read the saved key frame pose and key frame feature point cloud;

[0048] Read the unreliable GPS position information at the current moment gps .x vague ,pose gps .y vague ,heading vague ) as coarse positioning information; (pose gps .x,pose gps.y) is the two-dimensional map location coordinate information provided by GPS, heading gps It is the attitude information provided by GPS;

[0049] Searching for a key frame pose whose distance to the coarse positioning information is within a third set threshold value and a key frame feature point cloud corresponding to the key frame pose in the read key frame pose sequence;

[0050] According to the key frame pose, the key frame feature point cloud is spliced ​​to obtain two local feature point cloud maps, namely the plane feature point cloud map flatMap and the edge line feature point cloud map cornerMap;

[0051] Get the plane line feature point cloud and edge line feature point cloud in the current frame point cloud;

[0052] Find the edge line feature point Corner in the edge line feature point cloud map cornerMap i The two closest points, calculate the Corner i The distance to the straight line formed by these two points D_corner i ; Flatten the plane feature points in the current frame point cloud i , find the plane feature point Flat in the plane feature point cloud map flatMap i The three nearest points, calculate Flat i The distance to the plane formed by these three points is D_flat i ;

[0053] The n edge line feature points of the current frame correspond to the distance sequence {D_corner1, D_corner2, ..., D_corner n}, construct the objective function:

[0054]

[0055] Get F corner The minimum posture transformation relationship [t x ,t y ,θ yaw ], t x It is the displacement change of the AGV center in the x-direction in the map coordinate system corresponding to the established prior global point cloud map. The origin of the map coordinate system is the starting point of the vehicle at the initial moment of map construction. The x-axis of the map coordinate system is defined as the longitudinal center axis passing through the center of the rear axle of the vehicle at the initial moment. The direction from the center of the rear axle of the vehicle to the front of the vehicle along the longitudinal center axis is the positive direction of the x-axis. The y-axis of the map coordinate system is defined as passing through the center of the rear axle of the vehicle at the initial moment and perpendicular to the longitudinal center axis of the vehicle. The direction from the center of the rear axle of the vehicle to the left of the front of the vehicle is the positive direction of the y-axis. yis the displacement change of the AGV center in the y direction of the map coordinate system, θ yaw is the change in AGV yaw angle;

[0056] Use the following formula to obtain the initial pose description in the map coordinate system:

[0057] (pose lidar .x,pose lidar .y,heading lidar ):

[0058]

[0059] Complete positioning initialization pose.x = pose lidar .x,pose.y=pose lidar .y,heading=heading lidar ; q0, q1, q2, q3 are quaternions describing posture changes; T cx T c The displacement change in the x direction, T cy T c The change in position in the y direction; T c =T v T, T is the rotation matrix from the rough positioning pose to the correct initial pose, T v It is the pose rotation matrix from the origin of the global point cloud map to the coarse positioning pose.

[0060] The above process extracts edge feature information from the environmental point cloud to obtain the vehicle's displacement in the x and y directions and the change in yaw angle. This eliminates the need to align the entire point cloud, reduces the amount of computation, and ensures real-time positioning. At the same time, matching based on plane features provides constraints for the entire alignment process, avoiding mismatches based solely on edge features, and ensuring the robustness and accuracy of positioning.

[0061] In step S3, during the positioning and navigation process based on the laser radar point cloud registration, if the GPS signal becomes available again and the difference between the laser radar pose and the GPS pose is less than a given threshold, the specific implementation process of switching to GPS positioning includes:

[0062] If d <d max And h <h max , it is considered that the current state switching will not cause posture jump, and the positioning mode is switched to GPS-based positioning; otherwise, it is considered that the positioning mode switching in the current state will affect the execution of the navigation function, and the posture information provided by the lidar will continue to be used for navigation; among them,

[0063]

[0064] h=heading-heading gps ;

[0065] (pose gps .x,pose gps .y,heading gps ) is the position provided by GPS; (pose.x, pose.y, heading) is the current AGV position; (pose gps .x,pose gps .y) is the two-dimensional map location coordinate information provided by GPS, heading gps It is the attitude information provided by GPS.

[0066] During the GPS-based positioning and navigation process, if the GPS availability flag changes from true to false, and the difference between the position based on the lidar point cloud registration and the GPS position at the previous moment is less than a given threshold, the specific implementation process of switching to the positioning based on the lidar point cloud registration includes:

[0067] If d lidar <d max And h lidar <h max , it is considered that the current state switching will not cause posture jump, and the positioning mode is switched to positioning based on lidar point cloud registration; otherwise, it is considered that the positioning mode switching in the current state will affect the execution of the navigation function, and the GPS signal is waiting to return to normal. If the GPS signal has not returned to normal after more than T1 minutes, an abnormal report instruction is sent to the user terminal, and the user performs manual intervention; among them,

[0068]

[0069]

[0070] is the GPS position at the last moment; lidar .x, pose lidar .y, heading lidar ) is the current pose based on the LiDAR point cloud registration;

[0071] After determining that the current GPS signal is in an available state, the present invention does not immediately perform positioning switching, but instead compares the posture provided by the matching positioning based on the prior map with the posture provided by the GPS. The switching is performed only when the difference between the two postures is within the set threshold range. This can further ensure the availability of the GPS signal and avoid the adverse effects of positioning jumps caused by rash switching on the navigation planning function.

[0072] The present invention also provides a computer device, comprising a memory, a processor and a computer program stored in the memory; the processor executes the computer program to implement the steps of the method of the present invention.

[0073] The present invention also provides a computer program product, comprising a computer program / instruction; when the computer program / instruction is executed by a processor, the steps of the method of the present invention are implemented.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] 1) The present invention overcomes the positioning problem of AGV when GPS signals are unavailable in outdoor working scenarios, ensuring that AGV is not affected by weather and obstructed areas (such as tunnels and greenhouses in farms) when performing navigation tasks, and ensuring the stability of the continuous domain of navigation.

[0076] 2) By constructing the triple criteria for GPS signal availability, it provides a strong guarantee for the accurate judgment of GPS signal availability and the validity of navigation and positioning information;

[0077] 3) The stored key frame point cloud data and key frame pose information are matched with the point cloud data obtained by the real-time lidar to obtain the positioning information based on the lidar, thus avoiding the loss of positioning information when the GPS signal is unavailable;

[0078] 4) Through the design of two positioning initialization schemes, it is ensured that AGV can complete positioning initialization under any conditions; according to the availability judgment results of GPS signals and laser radar-based positioning information, a positioning mode selection and switching mechanism in the AGV navigation operation process is proposed to ensure the smooth and stable execution of navigation operation tasks.

[0079] 5) Compared with the existing positioning switching method, the present invention adopts the established prior point cloud map for matching positioning in the laser radar-based positioning part. The prior map is constructed based on the GPS signal, which avoids the influence of the cumulative error generated over time; and the method of the present invention calibrates the occluded area of ​​the prior map, and designs a reliability judgment basis based on the continuity of GPS positioning data, which provides a more reliable basis for switching positioning schemes. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 This is a flow chart of a method according to an embodiment of the present invention;

[0081] Figure 2 This is a flow chart of determining the availability of GPS signals according to the present invention;

[0082] Figure 3 A flowchart of the present invention for building a map based on collected data;

[0083] Figure 4 Comparison chart of positioning trajectories provided for outdoor GPS and LiDAR point cloud matching;

[0084] Figure 5 It is the error curve of the GPS true value coordinates within 30 seconds after the positioning mode is switched from GPS-based positioning to laser radar point cloud matching-based positioning;

[0085] Figure 6 It is the difference curve of the position coordinates based on the lidar point cloud registration within 5 seconds after the positioning mode is switched back to the GPS-based positioning mode and before the switching. DETAILED DESCRIPTION

[0086] The present invention implements an AGV outdoor positioning switching method and system, comprising the following steps:

[0087] Step 1: Manually control the AGV equipped with a combined navigation system and a multi-line laser radar to collect data on the work scene, and build a map and save features of the work scene based on the parsed GPS data and laser radar point cloud data;

[0088] Step 2: calibrate the GPS signal unreliable area in the established work scene prior global point cloud map, and expand the calibrated GPS unreliable area according to the boundary points to obtain the position representation method of the GPS unreliable area in the navigation coordinate system;

[0089] Step 3, based on the reliability original flag A of GPS stable positioning and orientation, the GPS signal reliability flag B determined by the current posture and the position of the calibrated GPS signal shielding area, and the GPS signal reliability flag C determined by the continuity judgment of the GPS posture information at the previous and next moments, jointly determine the availability status of the current GPS signal;

[0090] Step 4: Initialize the outdoor positioning of the AGV based on the laser radar and GPS. At the initial moment, the availability of the GPS signal is judged, and the corresponding positioning initialization mode is selected according to the judgment result to obtain the initial posture of the AGV in the outdoor scene.

[0091] Step 5: Select and switch the positioning scheme during the AGV navigation operation according to the positioning initialization mode and the results of the real-time availability judgment of the GPS signal and the laser radar positioning information;

[0092] The specific implementation process of each step is as follows:

[0093] Step 1 is to manually control the AGV equipped with a combined navigation system and a multi-line laser radar to collect data on the work scene, and to map and save features of the work scene based on the parsed GPS data and laser radar point cloud data.

[0094] Step 1.1, parse the GPS raw data;

[0095] Step 1.1.1, read the GPS raw data from the serial port, parse it according to the data parsing standard given by the manufacturer (different manufacturers have different data protocol specifications), and get the current geographic location information (Latitude, Longitude, Altitude);

[0096] Step 1.1.2, select the GPS coordinate origin and record its latitude and longitude information (Latitude0, Longitude0, Altitude0);

[0097] Step 1.1.3: Reset the coordinate origin based on the Geographiclib library and the latitude and longitude information of the selected GPS coordinate origin, and obtain the pose information of the current position relative to the coordinate origin based on the geo_converter class in the library function. gps .x,pose gps .y,pose gps .z,heading gps ,pitch gps ,roll gps );

[0098] Step 1.2: Control the AGV equipped with a multi-line laser radar and an integrated navigation system to collect data in a specified working scene;

[0099] Step 1.2.1, manually control the AGV equipped with lidar and GPS to start data collection of the working scene during a period of clear weather and with few pedestrians and vehicles. The collected information includes the pose information provided by GPS, which is recorded in the nav_msgs::Odometry data format in ROS, and the point cloud data provided by the lidar, which is recorded in the sensor_msgs::PointCloud2 data format in ROS. All data are saved in the form of rosbag;

[0100] Step 1.2.2, control the vehicle to collect data at a speed less than 0.5m / s, and try to drive in a straight line and close to the side with obvious environmental features during driving;

[0101] Step 1.2.3: When the vehicle is about to enter a blocked area in the work scene (the working area between ridges in the shed, dense vegetation, tunnels, etc.), a signal is sent to the on-board industrial computer through a wireless device. After receiving the signal, the on-board industrial computer records the timestamp of entering the blocked area in ROS. s ;

[0102] Step 1.2.4: Complete the data collection work in the signal blocking area. When you are about to leave the signal blocking area, send a signal to the on-board industrial computer through a wireless device. After receiving the signal, the on-board industrial computer records the timestamp t of leaving the blocking area in ROS. e ;

[0103] Step 1.2.5: After the vehicle completes data collection in all work areas, it returns to the starting position. After returning to the starting position, data recording is not stopped immediately. The AGV continues to move along the path it traveled at the initial moment for at least 10 seconds.

[0104] Step 1.3, replay the recorded rosbag format data set, and construct a global 3D point cloud map based on the SLAM method;

[0105] Step 1.3.1, select different pose sources according to the timestamp of the marked signal occlusion area to save the information required for mapping;

[0106] Step 1.3.1.1, if the current data timestamp does not belong to the time period of the marked signal occlusion, then the GPS pose information is reliable at this time, and the pose information gps_pose(pose gps .x,pose gps .y,pose gps .z,heading gps ,pitch gps ,roll gps ) is assigned to the SLAM algorithm based on the current frame point cloud and the local point cloud sub-graph built in real time to solve the pose lidar_maping_pose (pose lidar .x,pose lidar .y,pose lidar .z,heading lidar ,pitch lidar ,roll lidar )

[0107] Step 1.3.1.2, if the current data timestamp belongs to the time period of the marked signal occlusion, the absolute pose information obtained by GPS is used as the initial value of the laser odometer, and the assignment operation of gps_pose to lidar_mapping_pose is stopped. The point cloud registration based on the laser radar is completed in the occluded area for real-time pose estimation, and the pose is corrected through loop closure detection;

[0108] Step 1.3.2: Determine the moving distance based on the real-time acquired position information, and save the key frame position and key frame point cloud when the position movement is greater than 2m;

[0109] Step 1.3.3, stitching the map based on the saved key frame poses and key frame point clouds to obtain a global point cloud map;

[0110] Step 1.3.3.1, read the n key frame poses key_frame_pose_1, key_frame_pose_2, ..., key_frame_pose_n and n key frame point clouds key_frame_1, key_frame_2, ..., key_frame_n saved in step 1.3.2;

[0111] Step 1.3.3.2: Filter out ground points based on the angle between the corresponding points on two adjacent scanning lines of the multi-line laser radar and the ground. Filter out ground points for each key frame point cloud, but do not delete them completely. Save them as one of the features of the point cloud.

[0112] Step 1.3.3.3, based on the initial pose init_pose (i.e., the map origin) at the initial moment, import each keyframe point cloud data after filtering out the ground points into the global map according to its corresponding keyframe pose to obtain a spliced ​​dense point cloud map;

[0113] Step 1.3.3.4: Use the voxel grid filter in the PCL library to downsample and remove point cloud data with similar positions in space to obtain a sparse global point cloud map after removing the ground points;

[0114] Step 1.3.3.5, collapse the downsampled point cloud map into two dimensions to obtain a grid map that can be used for navigation tasks;

[0115] Step 1.3.4, extract features from the retained key frame point cloud to obtain corresponding plane feature maps and edge line feature maps;

[0116] Step 1.3.4.1, read the key frame point cloud obtained in step 1.3.3.2 after removing the ground points;

[0117] Step 1.3.4.2, project the point cloud data into an H*W image, where each point cloud corresponds to a pixel grid in the image, where H is the number of lines of the multi-line laser radar, and W can be obtained from the horizontal resolution δ of the multi-line laser radar:

[0118] W=360° / δ

[0119] For example, for the Velodyne-16 lidar, H = 16, δ = 0.2°, W = 360° / 0.2° = 1800;

[0120] Step 1.3.4.3: Cluster the projected point clouds according to their adjacent positional relationships in the image and the distance between their origin point clouds and the origin of the LiDAR coordinate system, and delete the point cloud clusters with less than 30 point clouds;

[0121] Step 1.3.4.4: Calculate the surface smoothness of the K point cloud clusters after clustering, and calculate the n points contained in each point cloud cluster. k Roughness around a point:

[0122]

[0123] Where S is the point p in the kth point cloud cluster ki Select a set of 5 points on each side of the row where the projection point on the image is located, r ki is the distance between the laser point in the kth point cloud cluster and the origin of the laser radar coordinate system, r kj is the point p in the k-th point cloud cluster ki The distance between the point cloud in the corresponding point set S and the origin of the laser radar coordinate system;

[0124] Step 1.3.4.5: The calculated roughness c ki With threshold c th For comparison, if c ki >c th It belongs to the edge feature. If c ki <c th It belongs to the plane feature;

[0125] Step 1.3.4.6, extract the points judged as plane features in the point cloud to form a plane feature point cloud, extract the points judged as edge features in the point cloud to form an edge feature point cloud, and save the key plane feature point cloud and edge feature point cloud corresponding to each key frame point cloud;

[0126] Step 1.3.4.7, based on the initial pose init_pose at the initial moment (i.e., the map origin), import all plane features into the global map according to their corresponding key frame poses to obtain a global plane feature map; import all edge features into the global map according to their corresponding key frame poses to obtain a global edge feature map;

[0127] Step 2 is to calibrate the GPS signal unreliable area in the established work scene prior global point cloud map, and expand the calibrated GPS unreliable area according to the boundary points to obtain the position representation of the GPS unreliable area in the navigation coordinate system;

[0128] Step 2.1, read the global point cloud map obtained in step 1.3.3.4;

[0129] Step 2.2, manually calibrate the m unreliable areas in the global point cloud map, mark the boundary points of the unreliable areas, and record the coordinate information of the boundary points in the map;

[0130] Step 2.3, dilate the signal occlusion area based on the boundary points;

[0131] Step 2.3.1: For the signal occlusion area 1, assume that n boundary points (n>3) are manually marked, and 3 points are randomly selected from the n boundary points to form a set of arrangements. indivual;

[0132] Step 2.3.2, calculate the radius of the circumscribed circle corresponding to each arrangement, and select the arrangement with the largest circumscribed circle radius as the marking boundary point of the signal blocking area;

[0133] Step 2.3.3, calculate the coordinates (x block1 ,y block1 ) and the radius of the outer circle r1, the expansion operation is performed on each signal-blocking area in the map to obtain the center coordinate sequence of the expansion area (x block1 ,y block1 ),(x block2 ,y block2 ),...,(x blockm ,y blockm ) and the expansion zone radius sequence r1,r2,...,r m ;

[0134] Step 3, based on the reliability original flag A of GPS stable positioning and orientation, the GPS signal reliability flag B determined by the current posture and the position of the calibrated GPS signal shielding area, and the GPS signal reliability flag C determined by the continuity judgment of the GPS posture information at the previous and next moments, jointly determine the availability status of the current GPS signal;

[0135] Step 3.1, read the GPS raw data, parse the GPS raw data, and obtain the current positioning status flag information. The flag information is true when the vehicle can successfully obtain its real-time position and posture. At this time, the GPS reliability flag A=1, otherwise the GPS reliability flag A=0

[0136] Step 3.2, obtain the GPS reliability flag B according to the positional relationship between the current posture and the signal unreliable area marked in the prior map;

[0137] Step 3.2.1, read the current position information of AGV in the map (pose.x, pose.y);

[0138] Step 3.2.2, read the signal occlusion expansion center coordinate sequence and expansion area radius sequence obtained in step 2.3.3;

[0139] Step 3.2.3: Calculate the distance d between the current position of the AGV and the expansion center of the signal occlusion area i in the expansion center coordinate sequence i :

[0140]

[0141] Step 3.2.4: If d i <r i , then the AGV is in the signal unreliable area, the reliability flag B=0, and the calculation of d is stopped i ;

[0142] Step 3.2.5: If d i >r i And if the traversal of the signal occlusion expansion center coordinate sequence is completed, the reliability flag B = 1, otherwise return to step 3.2.3;

[0143] Step 3.3, deriving the GPS signal reliability flag C according to the current AGV position information in the map, vehicle speed information and data jump threshold;

[0144] Step 3.3.1, obtain the current vehicle running linear speed v and steering angular speed w according to the wheel odometer information;

[0145] Step 3.3.2: Read the current AGV's position information in the map (pose.x, pose.y, heading) and the previous AGV's position information in the map (pose.x last ,pose.y last ,heading last );

[0146] Step 3.3.3: Calculate the position change threshold H d :

[0147] H d = k × v × T

[0148] Where T is the time interval for GPS signal release, k is the threshold expansion coefficient, which is adjustable from 1.2 to 1.5;

[0149] Step 3.3.4: Calculate the yaw angle change threshold H h :

[0150] H h = k × w × T

[0151] Where T is the time interval for GPS signal release, k is the threshold expansion coefficient, which is adjustable from 1.2 to 1.5;

[0152] Step 3.3.5, calculate the position difference d and posture difference h of the AGV between the current moment and the previous moment:

[0153]

[0154] h=heading-heading last

[0155] If d <H d And h <H h Then the GPS signal reliability flag C is set to 1, otherwise the GPS signal reliability flag C is set to 0;

[0156] Step 3.4: If the GPS reliability determination flags A=1, B=1, and C=1 are satisfied at the same time, the GPS availability flag is set to true, otherwise it is set to false;

[0157] Step 4: Initialize the outdoor positioning of the AGV based on the laser radar and GPS. At the initial moment, the availability of the GPS signal is judged, and the corresponding positioning initialization mode is selected according to the judgment result to obtain the initial posture of the AGV in the outdoor scene.

[0158] Step 4.1, after the AGV is started, read the judgment result of the GPS availability flag in step 3;

[0159] Step 4.2: If the GPS availability flag is true, use the GPS-provided pose to initialize the positioning pose.x = pose gps .x,pose.y=pose gps .y,heading=heading gps The positioning initialization process is completed;

[0160] Step 4.3: If the result of the GPS availability flag is false, use the laser radar matching positioning to complete the positioning initialization;

[0161] Step 4.3.1, read the key frame pose and key frame feature point cloud obtained in step 1;

[0162] Step 4.3.2: Read the unreliable GPS position information (pose gps .x vague ,pose gps .y vague ,heading vague ) as coarse positioning;

[0163] Step 4.3.3, find the key frame pose within 10m near the coarse positioning and its corresponding key frame feature point cloud in the read key frame pose sequence;

[0164] Step 4.3.4, stitch the key frame feature point cloud according to the key frame pose to obtain two local feature point cloud maps, the plane feature point cloud map flatMap and the edge line feature point cloud map cornerMap;

[0165] Step 4.3.5, execute steps 1.3.4.2 to 1.3.4.6 on the point cloud of the current frame to obtain the plane line feature point cloud and the edge feature point cloud in the current frame;

[0166] Step 4.3.6: Corner the edge feature point in the current frame point cloud i , find the two closest points in the edge feature point cloud map cornerMap (make sure the two points come from two different laser lines), and calculate Corner i The distance to the straight line formed by these two points D_corner i ;

[0167] Step 4.3.7: Flatten the plane feature points in the current frame point cloud i , find the three closest points in the flat feature point cloud map flatMap (make sure that the three points do not belong to the same laser line), and calculate Flat i The distance to the plane formed by these three points is D_flat i ;

[0168] Step 4.3.8, the n edge feature points of the current frame correspond to the distance sequence {D_corner1, D_corner2, ..., D_corner n}, construct the objective function:

[0169]

[0170] By solving a nonlinear optimization problem using the Levenberg-Marquardt method, we can obtain an optimal solution for F corner The minimum posture transformation relationship [t x ,t y ,θ yaw ], t x is the displacement change of the AGV center in the x-direction in the map coordinate system corresponding to the established prior global point cloud map. The origin of the coordinate system is the starting point of the vehicle at the initial moment of map construction. y is the displacement change of the AGV center in the y direction of the map coordinate system, θ yaw is the change in AGV yaw angle;

[0171] Step 4.3.9: Use the following formula to obtain the initial pose description in the map coordinate system (pose lidar .x,pose lidar .y,heading lidar ):

[0172]

[0173] Complete positioning initialization pose.x = pose lidar .x,pose.y=pose lidar .y,heading=heading lidar ; q0, q1, q2, q3 are quaternions describing posture changes; T cx T c The displacement change in the x direction, T cy T c The change in position in the y direction; T c =T v T, T is the rotation matrix from the rough positioning pose to the correct initial pose, T v It is the pose rotation matrix from the origin of the global point cloud map to the coarse positioning pose.

[0174] Step 4.3.10: Use the pose information obtained in step 4.3.10 to initialize positioning, pose.x = pose lidar .x,pose.y=pose lidar .y,heading=heading lidar

[0175] Step 5: Select and switch the positioning scheme during the AGV navigation operation according to the positioning initialization mode and the results of the real-time availability judgment of the GPS signal and the laser radar positioning information;

[0176] Step 5.1: After the positioning initialization process is completed, the AGV waits for the release of the navigation task;

[0177] Step 5.2: After the navigation task is released, the positioning mode for the initial stage of subsequent navigation tasks is selected according to the initialization mode used during positioning initialization;

[0178] Step 5.2.1, if the positioning initialization mode is initialization based on GPS posture information, the AGV continues to use the posture information provided by the GPS to perform the navigation task, and performs real-time signal detection and judgment on whether the navigation task is completed according to the GPS availability judgment process in step 3. If the navigation task is completed, the positioning module is turned off;

[0179] Step 5.2.1.1: If the GPS availability flag is detected to be false during the navigation task, the AGV will suspend the navigation task.

[0180] Step 5.2.1.2, the AGV waits for 30 seconds at the origin. If the GPS availability flag is restored to true and the duration exceeds 30 seconds, the AGV continues to use the position information provided by the GPS for navigation tasks. If the GPS availability flag is still false after the waiting time, the laser radar positioning availability is determined.

[0181] Step 5.2.1.3: Use the GPS position information of the last moment As the initial pose information for point cloud registration, execute according to the process of steps 4.3.1-4.3.10 to obtain the pose information based on lidar point cloud registration lidar (pose lidar .x,pose lidar .y,heading lidar );

[0182] Step 5.2.1.4, calculate the position difference d and the yaw angle difference h:

[0183]

[0184]

[0185] Step 5.2.1.5: The calculated position difference d lidar The position difference threshold d max Compare and convert the yaw angle difference h lidar The difference between the yaw angle and the threshold h max Compare, if d lidar <d max And h lidar <hmax The laser radar positioning availability flag position is true, otherwise it is set to false;

[0186] Step 5.2.1.5.1, if the LiDAR positioning availability flag is true, use the current pose information obtained based on point cloud registration as the prior information for searching the key frame pose and key frame feature point cloud in step 4.3.3, and execute the process of steps 4.3.1-4.3.10 to obtain the pose information based on point cloud registration at the new moment;

[0187] Step 5.2.1.5.2, if the LiDAR positioning availability flag is false, continue to wait for the GPS signal to return to normal. If the GPS signal still does not return to normal after more than 15 minutes (T1), send an abnormal report command to the user terminal for manual intervention by the user;

[0188] Step 5.2.2: If the positioning initialization mode is initialization based on laser radar matching positioning, select laser radar positioning as the positioning information source for the navigation task execution;

[0189] Step 5.2.2.1: Assign the current position information provided by the laser radar positioning to the position storage variable And use it as the initial pose information for point cloud registration;

[0190] Step 5.2.2.2: Follow the process of steps 4.3.1-4.3.10 to obtain the pose information based on the laser radar point cloud registration at the new moment. lidar (pose lidar .x,pose lidar .y,heading lidar );

[0191] Step 5.2.2.3: According to the GPS availability judgment process in step 3, the signal detection is performed in real time and the judgment of whether the navigation task is completed is judged. If the navigation task is completed, the positioning system is shut down;

[0192] If the GPS availability flag is detected to be true during the navigation task, the pose provided by GPS is gps .x,pose gps .y,heading gps ) is compared with the current pose (pose.x, pose.y, heading) to calculate the position difference d and the yaw angle difference h:

[0193]

[0194] h=heading-headinggps

[0195] Step 5.2.2.4: Compare the calculated position difference d with the position difference threshold d max Compare the yaw angle difference h with the yaw angle difference threshold h max Compare, if d <d max And h <h max It is believed that the current state switching will not cause posture jump, will not affect the normal operation and stability of AGV planning and control, and the positioning mode is switched to GPS-based positioning; otherwise, it is believed that the positioning mode switching in the current state will affect the execution of the navigation function, and the posture information provided by the lidar will continue to be used for navigation.

[0196] The location information provided by GPS and point cloud registration in a trajectory is recorded in an outdoor environment, which contains 8832 sets of coordinate data. Some of the data are recorded in Table 1, and the visualization results are as follows: Figure 4 As shown in the figure, the dotted line part is the GPS positioning trajectory, and the solid line part is the positioning trajectory based on lidar point cloud registration. The positioning error based on point cloud registration and the positioning based on combined navigation does not exceed 10cm, which proves the feasibility and reliability of lidar as a switching alternative.

[0197] Table 1 Positioning coordinates based on point cloud registration and GPS (excerpt)

[0198] Positioning coordinates (x, y) / (m, m) obtained based on point cloud registration Positioning coordinates obtained based on GPS (x, y) / (m, m) (14.3864212036,-60.009727478) (14.6258982895,-60.0205248978) (14.3467760086,-61.1161613464) (14.6013512498,61.0777746375) (14.3234176636,-62.1469535828) (14.5866234915,-62.1128597022) (14.3094949722,-63.1125450134) (14.577787321,-63.1191309194) (14.2816371918,-64.1092300415) (14.5817162362,-64.0832894497) … … (14.1234521866,-75.0794677734) (14.5542356198,-75.099078066) (14.120757103,-76.0172805786) (14.521832858,-76.005608644) (14.0550909042,-77.2397613525) (14.5051414361,-77.2302006878) (14.0451145172,-78.134437561) (14.5041605748,-78.1311900775) (14.0101280212,-79.1993026733) (14.5031799617,-79.1973057105)

[0199] During outdoor navigation, the GPS signal is manually shielded and turned on to simulate the conditions for triggering positioning switching, so that the AGV completes the switch from GPS-based positioning to laser radar point cloud registration-based positioning, and switches back to GPS-based positioning mode after driving a certain distance, and repeats such simulation tests. Here, a set of test data is selected for display, and the error curve of the true value of the GPS position within 30 seconds after switching to laser radar point cloud registration-based positioning is plotted, as well as the difference curve of the positioning position before switching to laser radar point cloud registration-based positioning within 5 seconds after switching back to GPS-based positioning, as shown in the figure. Figure 5 , Figure 6 It can be seen that the position error at the time of the two switches can be basically controlled within 10 cm, and there will be no posture jump to affect the navigation function.

[0200] Table 2 contains the navigation positioning coordinates and GPS true value coordinates of the two positioning switches (excerpt)

[0201] Positioning coordinates used during navigation (x,y) / (m,m) Based on the true coordinates provided by GPS (x, y) / (m, m) (23.5787726874，-65.8486973888) (23.5787726874，-65.8486973888) (23.5777908288，-65.9107581696) (23.5777908288，-65.9107581696) (23.5768090487，-66.0359879579) (23.5768090487，-66.0359879579) … … (23.5905635375，-72.0514508357) (23.5905635375，-72.0514508357) (23.6102035613，-73.1585706658) (23.6102035613，-73.1585706658) (Positioning switched from GPS to LiDAR) (Positioning switched from GPS to LiDAR) (23.1712560654，-73.4919281006) (23.6141317374，-73.5154201405) (23.1480178833，-74.1129226685) (23.5915479369，-74.0817247755) (23.1380710602，-74.9416046143) (23.5611089526，-74.9273026886) … … (23.9201078415，-83.214225769) (23.4658712963，-83.2223903636) (23.8575553894，-84.0822296143) (23.4334685071，-84.0923496495) (23.8493518829，-84.9674911499) (23.4167766363，-84.9623086508) (23.8528795242，-85.2276992798) (23.4089214894，-85.2227424147) (Positioning switched from LiDAR to GPS) (Positioning switched from LiDAR to GPS) (23.4079396293，-85.2848031931) (23.4079396293，-85.2848031931) (23.4030302353，-85.5186393428) (23.4030302353，-85.5186393428) (23.3853564012，-86.369758463) (23.3853564012，-86.369758463) … … (23.351978714，-93.1487898792) (23.351978714，-93.1487898792) (23.3205575379，-93.7793718845) (23.3205575379，-93.7793718845) (23.3077928661，-94.1949574575) (23.3077928661，-94.1949574575)

Claims

1. An AGV outdoor positioning switching method, characterized in that: The following steps are involved: S1. Collect AGV working scene data and establish a priori global point cloud map of the working scene; S2. constructing a representation method of an unreliable GPS signal area using the prior global point cloud map of the working scene to determine the availability status of the current GPS signal; S3. If the current GPS signal is unavailable, the availability of the laser radar positioning information is determined. If the laser radar positioning information is available, the laser radar positioning is switched to; if the laser radar is used to complete the positioning initialization, when the GPS signal is available and the difference between the laser radar posture and the GPS posture is less than a given threshold, the laser radar positioning is switched to the GPS positioning; The specific implementation process of step S1 includes: Control the AGV to move forward at a speed less than the first set threshold, and record the timestamp t of the AGV entering the blocked area s And the timestamp t of the AGV leaving the blocked area e ; Different mapping pose acquisition methods are selected according to the timestamp of the marked signal-blocked area. If the current data is not in the signal-blocked area, the pose information provided by GPS is used for mapping; if the current data belongs to the signal-blocked area, the pose information provided by point cloud matching is used for mapping. Determine the moving distance of the vehicle according to the posture information obtained in real time, and save the key frame posture and key frame point cloud when the moving distance is greater than a second set threshold; Splice the key frame poses and key frame point clouds to obtain a global point cloud map; In step S2, the specific implementation process of constructing a representation method of the GPS signal unreliable area using the prior global point cloud map of the working scene includes: Read the global point cloud map, calibrate the m signal blocking areas in the global point cloud map, mark the boundary points of the signal blocking areas, and record the coordinate information of the boundary points in the map; Dilate the signal occlusion area based on the boundary points; For any signal blocking area, suppose n boundary points are marked, and 3 points are randomly selected from the n boundary points to form a set of arrangements. The radius of the circumscribed circle corresponding to each arrangement is calculated, and the arrangement with the largest circumscribed circle radius is selected as the marking boundary point of the signal blocking area; n>3; Calculate the coordinates (x block1 ,y block1 ) and the outer circle radius r1; The above expansion operation is performed on all signal occlusion areas in the global point cloud map to obtain the center coordinate sequence of the expansion area (x block1 ,y block1 ),(x block2 ,y block2 ),...,(x blockm ,y blockm ) and the expansion zone radius sequence r1,r2,...,r m ,Each expansion center coordinate plus the expansion area radius corresponding to the expansion center represents a circular GPS signal unreliable area in the prior map; In step S2, the specific implementation process of determining the available status of the current GPS signal includes: Read the GPS raw data, parse the GPS raw data, and obtain the current positioning status flag information. The flag information is true when the vehicle can obtain its own real-time position and posture. At this time, the GPS reliability flag A=1, otherwise the GPS reliability flag A=0; According to the positional relationship between the current posture and the calibrated GPS signal unreliable area, the GPS reliability flag B is obtained; According to the current AGV position information in the map, vehicle speed information and data jump threshold, the GPS signal reliability flag C is obtained; If the GPS reliability determination flags A=1, B=1, and C=1 are satisfied at the same time, the GPS availability flag is set to true, that is, the GPS signal is available; otherwise, it is set to false, that is, the GPS signal is unavailable; In step S3, the specific implementation process of using laser radar to complete positioning initialization includes: Read the saved key frame pose and key frame feature point cloud; Read the unreliable GPS position information at the current moment As rough positioning information; (pose gps .x,pose gps .y) is the two-dimensional map location coordinate information provided by GPS, heading gps It is the attitude information provided by GPS; Searching for a key frame pose whose distance to the coarse positioning information is within a third set threshold value and a key frame feature point cloud corresponding to the key frame pose in the read key frame pose sequence; According to the key frame pose, the key frame feature point cloud is spliced ​​to obtain two local feature point cloud maps, namely the plane feature point cloud map flatMap and the edge line feature point cloud map cornerMap; the plane line feature point cloud and the edge line feature point cloud in the current frame point cloud are obtained; Find the edge line feature point Corner in the edge line feature point cloud map cornerMap i The two closest points, calculate the Corner i The distance to the straight line formed by these two points D_corner i ; Flatten the plane feature points in the current frame point cloud i , find the plane feature point Flat in the plane feature point cloud map flatMap i The three nearest points, calculate Flat i The distance to the plane formed by these three points is D_flat i ; The n edge line feature points of the current frame correspond to the distance sequence {D_corner1, D_corner2, ..., D_corner n }, construct the objective function: Get F corner The minimum posture transformation relationship [t x ,t y ,θ yaw ]; t x It is the displacement change of the AGV center in the x-direction in the map coordinate system corresponding to the established prior global point cloud map. The origin of the map coordinate system is the starting point of the vehicle at the initial moment of map construction. The x-axis of the map coordinate system is defined as the longitudinal center axis passing through the center of the rear axle of the vehicle at the initial moment. The direction from the center of the rear axle of the vehicle to the front of the vehicle along the longitudinal center axis is the positive direction of the x-axis. The y-axis of the map coordinate system is defined as passing through the center of the rear axle of the vehicle at the initial moment and perpendicular to the longitudinal center axis of the vehicle. The direction from the center of the rear axle of the vehicle to the left of the front of the vehicle is the positive direction of the y-axis. y is the displacement change of the AGV center in the y direction of the map coordinate system, θ yaw is the change in AGV yaw angle; Use the following formula to obtain the initial pose description in the map coordinate system (pose lidar .x,pose lidar .y,heading lidar ): Complete positioning initialization pose.x = pose lidar .x,pose.y=pose lidar .y,heading=heading lidar ; q0, q1, q2, q3 are quaternions describing posture changes; T cx T c The displacement change in the x direction, T cy T c The position change in the y direction; T c =T v T, T is the rotation matrix from the rough positioning pose to the correct initial pose, T v It is the pose rotation matrix from the origin of the global point cloud map to the coarse positioning pose.

2. The AGV outdoor positioning switching method according to claim 1, characterized in that: According to the positional relationship between the current posture and the calibrated GPS signal unreliable area, the specific implementation process of obtaining the GPS reliability flag B includes: 1) Read the AGV's current position information (pose.x, pose.y) in the global point cloud map, the signal occlusion expansion center coordinate sequence, and the expansion area radius sequence r i , where i = 1, 2, 3, ..., m; 2) Calculate the distance d between the current position of the AGV and the expansion center of the signal occlusion area i in the expansion center coordinate sequence i ; 3) If d i <r i , then the AGV is in the signal unreliable area, the reliability flag B = 0, and the calculation of d is stopped i If d i >r i And the signal occlusion expansion center coordinate sequence traversal is completed, then the reliability flag B = 1; if d i >r i However, if the signal occlusion expansion center coordinate sequence is not traversed, then return to step 2).

3. The AGV outdoor positioning switching method according to claim 2, characterized in that: The specific implementation process of obtaining the GPS signal reliability flag C based on the current AGV position information in the map, vehicle speed information and data jump threshold includes: Calculate the current linear velocity v and steering angular velocity w of the AGV; Read the current AGV's position information in the map (pose.x, pose.y, heading) and the previous AGV's position information in the map (pose.x last ,pose.y last ,heading last );Calculate the position change threshold H d :H d = k × v × T; where T is the time interval for GPS signal release, and k is the threshold expansion coefficient; Calculate the yaw angle change threshold H h :H h = k × w × T; Calculate the position difference d and posture difference h of the AGV between the current moment and the previous moment: h=heading-heading last ; If d <H d And h <H h , then the GPS signal reliability flag C is set to 1, otherwise the GPS signal reliability flag C is set to 0; if the GPS reliability judgment flags A=1, B=1, and C=1 are satisfied at the same time, then the GPS availability flag is set to true, otherwise it is set to false.

4. The AGV outdoor positioning switching method according to claim 1, characterized in that: In step S3, during the positioning and navigation process based on the laser radar point cloud registration, if the GPS signal becomes available again and the difference between the laser radar pose and the GPS pose is less than a given threshold, the specific implementation process of switching to GPS positioning includes: If d <d max And h <h max , it is considered that the current state switching will not cause posture jump, and the positioning mode is switched to GPS-based positioning; otherwise, it is considered that the positioning mode switching in the current state will affect the execution of the navigation function, and the posture information provided by the lidar will continue to be used for navigation; among them, h=heading-heading gps ; (pose gps .x,pose gps .y,heading gps ) is the position provided by GPS; (pose.x, pose.y, heading) is the current AGV position; (pose gps .x,pose gps .y) is the two-dimensional map location coordinate information provided by GPS, heading gps It is the attitude information provided by GPS; During the GPS-based positioning and navigation process, if the GPS availability flag changes from true to false, and the difference between the position based on the lidar point cloud registration and the GPS position at the previous moment is less than a given threshold, the specific implementation process of switching to the positioning based on the lidar point cloud registration includes: If d lidar <d max And h lidar <h max , it is considered that the current state switching will not cause posture jump, and the positioning mode is switched to positioning based on lidar point cloud registration; otherwise, it is considered that the positioning mode switching in the current state will affect the execution of the navigation function, and the GPS signal is waiting to return to normal. If the GPS signal has not returned to normal after more than T1 minutes, an abnormal report instruction is sent to the user terminal, and the user performs manual intervention; among them, is the GPS position at the last moment; lidar .x, pose lidar .y, heading lidar ) is the current pose based on the lidar point cloud registration.

5. A computer device comprising a memory, a processor and a computer program stored in the memory; characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 4.

6. A computer program product comprising a computer program / instructions; characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images