Navigation method, medium and unmanned forklift truck

Abstract

The application relates to a navigation method and medium of an unmanned forklift and the unmanned forklift. The method comprises the following steps: processing a plurality of first point cloud data collected by a laser radar arranged on the unmanned forklift located in the compartment, and obtaining a plurality of second point cloud data; processing the plurality of second point cloud data, and extracting a plurality of third point cloud data from the plurality of second point cloud data; determining fourth point cloud data from the plurality of third point cloud data; processing the fourth point cloud data, generating a driving path of the unmanned forklift, and controlling the unmanned forklift to drive according to the driving path.

Description

Technical Field

[0001] This application relates to the field of unmanned forklift technology, specifically to a navigation method, medium, and unmanned forklift. Background Technology

[0002] In related technologies, when unmanned forklifts are picking up and placing goods in the cargo compartment of a box truck, the travel path of the unmanned forklift inside the compartment is uncontrollable due to the long length of the compartment, poor lighting conditions inside the compartment, and swaying during travel. Furthermore, the unmanned forklift cannot use QR code recognition technology to determine the position of the inner wall of the compartment and the goods in front, which reduces the operating efficiency of the unmanned forklift. Summary of the Invention

[0003] Therefore, it is necessary to provide a navigation method, medium, and unmanned forklift that can solve the problem of identifying the location of goods inside the forklift compartment, in order to address the above-mentioned technical issues.

[0004] A navigation method for an unmanned forklift includes the following steps:

[0005] Multiple first point cloud data collected by the lidar configured on the unmanned forklift located inside the truck body are processed to obtain multiple second point cloud data; the second point cloud data represent the point cloud data to be processed.

[0006] Multiple second point cloud data are processed to extract multiple third point cloud data; the third point cloud data represents the point cloud data located inside the carriage in the carriage center coordinate system;

[0007] Among multiple third-point cloud data, a fourth-point cloud data is determined; the fourth-point cloud data represents the data of the target cargo in the laser coordinate system.

[0008] The fourth point cloud data is processed to generate the driving path of the unmanned forklift, and the unmanned forklift is controlled to travel according to the driving path.

[0009] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the navigation method for the unmanned forklift described above.

[0010] An unmanned forklift includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the navigation method of the unmanned forklift described above.

[0011] The aforementioned navigation method, medium, and unmanned forklift involve processing multiple first point cloud data collected by a lidar system located inside the forklift compartment to obtain multiple second point cloud data. This second point cloud data is then processed to extract multiple third point cloud data. This process filters the point cloud data to identify the data located inside the compartment. From the third point cloud data, a fourth point cloud data is determined. This fourth point cloud data is then processed to generate the forklift's travel path. The forklift is then controlled to follow this path. This method enables the identification of goods positions within the point cloud data inside the compartment and generates the forklift's travel path, preventing collisions with the compartment walls and goods during travel. This effectively improves the forklift's efficiency and enables the successful handling of picking up and placing goods. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating the navigation method for an unmanned forklift in one embodiment;

[0013] Figure 2 This is a flowchart illustrating the navigation method for an unmanned forklift in yet another embodiment;

[0014] Figure 3 This is a flowchart illustrating the navigation method for an unmanned forklift in yet another embodiment;

[0015] Figure 4 This is a flowchart illustrating the navigation method for an unmanned forklift in yet another embodiment;

[0016] Figure 5 This is a flowchart illustrating the navigation method for an unmanned forklift in yet another embodiment;

[0017] Figure 6 This is a schematic diagram of cells divided on the YZ plane in one embodiment;

[0018] Figure 7 This is a schematic diagram of cells divided on the XZ plane in one embodiment;

[0019] Figure 8 This is a flowchart illustrating the navigation method for an unmanned forklift in one embodiment;

[0020] Figure 9 This is a flowchart illustrating the navigation method for an unmanned forklift in yet another embodiment;

[0021] Figure 10 This is a flowchart illustrating the navigation method for an unmanned forklift in yet another embodiment;

[0022] Figure 11 This is a flowchart illustrating the process of determining the driving path of an unmanned forklift in one application embodiment. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0024] The implementation details of the technical solutions in the embodiments of this application are described in detail below.

[0025] In one embodiment, such as Figure 1 As shown, a navigation method for an unmanned forklift is provided, which may include the following steps:

[0026] Step S101: Process the multiple first point cloud data collected by the lidar configured on the unmanned forklift located inside the vehicle compartment to obtain multiple second point cloud data.

[0027] The lidar is installed on the unmanned forklift. The specific installation location can be determined according to actual needs. For example, the lidar can be installed at the midpoint between the roots of the forks at both ends of the unmanned forklift.

[0028] The unmanned forklift operates inside the truck bed. When the forklift reaches a monitoring point inside the truck bed, its lidar is activated. The lidar emits laser signals to the surrounding area and collects the reflected laser signals, acquiring multiple first point cloud data sets. These first point cloud data sets are datasets of spatial points obtained by the lidar scanning. Each first point cloud data set contains three-dimensional coordinate information, namely X, Y, and Z elements. Based on this, the first point cloud data set represents the dataset of spatial points where the unmanned forklift is currently located. The space where the unmanned forklift is currently located includes the unmanned forklift, the goods, and the truck bed. In other words, the multiple first point cloud data sets can contain point cloud data belonging to the unmanned forklift, point cloud data belonging to the goods, and point cloud data belonging to the truck bed.

[0029] During the operation of an unmanned forklift, it is necessary to identify the inner wall of the truck bed and the location of the goods. Based on this, it is necessary to clean and filter the multiple first point cloud data acquired by the LiDAR to obtain multiple second point cloud data. The second point cloud data is the point cloud data that needs to be focused on. The second point cloud data contains the point cloud data of the unmanned forklift belonging to the goods and the point cloud data of the truck bed. Processing multiple second point cloud data can further determine the location of the inner wall of the truck bed and the goods.

[0030] Step S102: Process multiple second point cloud data and extract multiple third point cloud data from the multiple second point cloud data.

[0031] The second point cloud data includes point cloud data belonging to the walls of the truck bed (including the left wall, right wall, and bottom of the truck bed) and point cloud data belonging to the cargo. In order for the unmanned forklift to identify the location of the cargo, multiple second point cloud data need to be processed to extract multiple third point cloud data. The third point cloud data is the point cloud data located inside the truck bed, that is, the point cloud data in the truck bed excluding the left wall, right wall, and bottom of the truck bed. It can be understood that the cargo is stored inside the truck bed. Based on this, the third point cloud data can be regarded as the point cloud data belonging to the storage space of the truck bed. That is, the third point cloud data contains the point cloud data of the cargo. In this embodiment, the multiple third point cloud data are three-dimensional coordinate information in the truck bed center coordinate system, where the truck bed center coordinate system is a coordinate system established based on the center point of the truck bed.

[0032] In practical applications, multiple third point cloud data can be extracted by filtering multiple second point cloud data layer by layer. Specifically, a planar model of the left wall of the carriage can be constructed first from multiple second point cloud data. The point cloud data located outside the left wall of the carriage can be determined using the planar model of the left wall of the carriage. Then, a planar model of the right wall of the carriage can be constructed using the point cloud data located outside the left wall of the carriage. The point cloud data located outside the right wall of the carriage can be determined using the planar model of the right wall of the carriage (this part of the point cloud data is located outside both the left inner wall and the right wall of the carriage). Then, a ground model inside the carriage can be constructed using the point cloud data located outside the right wall of the carriage. The point cloud data located outside the ground inside the carriage can be determined using the ground model inside the carriage (this part of the point cloud data is located outside both sides of the carriage wall and the inner ground, which is the third point cloud data). The process of constructing the planar model of the left wall of the carriage, the planar model of the right wall of the carriage, and the model of the ground inside the carriage, as well as the process of determining the point cloud data located outside the left wall of the carriage, outside the right wall of the carriage, and outside the ground inside the carriage, will be explained in detail in the following content.

[0033] In one embodiment, such as Figure 2 As shown, multiple second-point cloud data are processed to extract multiple third-point cloud data, including:

[0034] Step S201: Based on multiple second point cloud data, determine the model parameters of the carriage center plane model and multiple fifth point cloud data.

[0035] Here, the model parameters of the carriage center plane model include A_middle, B_middle, C_middle, and D_middle. A_middle corresponds to the X-axis parameter, B_middle to the Y-axis parameter, C_middle to the Z-axis parameter, and D_middle to the constant term. This allows us to obtain the plane equation corresponding to the carriage center plane model, which can be expressed as: A_middle*x + B_middle*y + C_middle*z = D_middle, where (x, y, z) are different coordinate values. In practical applications, the model parameters of the carriage center plane model are obtained by processing multiple second point cloud data. Specifically, based on multiple second point cloud data, the model parameters of the carriage left wall plane model, the carriage right wall plane model, and the carriage interior ground model can be obtained, and then the model parameters of the carriage plane model can be derived.

[0036] In practical applications, processing multiple second point cloud data can also yield multiple fifth point cloud data. The fifth point cloud data represents the point cloud data located outside the side walls of the carriage and the interior ground. By processing multiple second point cloud data, the point cloud data located outside the side walls of the carriage and the interior ground can be separated, thereby obtaining multiple fifth point cloud data.

[0037] In one embodiment, such as Figure 3 As shown, based on multiple second point cloud data, the model parameters of the carriage center plane model are determined, including:

[0038] Step S301: Extract multiple sixth point cloud data from multiple second point cloud data, and process the multiple sixth point cloud data according to the set data processing flow to determine the model parameters of the left wall plane model of the carriage.

[0039] Here, multiple sixth-point cloud data are used to fit the point cloud data of the left wall of the carriage. That is, a planar model of the left wall of the carriage can be constructed using multiple sixth-point cloud data. In this embodiment, the multiple sixth-point cloud data are the point cloud data with x < 0 in multiple second-point cloud data. That is, the point cloud data with x < 0 in the second-point cloud data are extracted and determined as the sixth-point cloud data. Then, the multiple sixth-point cloud data are processed according to the set data processing flow to determine the model parameters of the planar model of the left wall of the carriage. In this embodiment, the model parameters of the planar model of the left wall of the carriage include A_left, B_left, C_left and D_left, where A_left corresponds to the X-axis parameter of the planar model of the left wall of the carriage, B_left corresponds to the Y-axis parameter of the planar model of the left wall of the carriage, C_left corresponds to the Z-axis parameter of the planar model of the left wall of the carriage, and D_left corresponds to the constant term of the planar model of the left wall of the carriage. The set data processing flow here is the data processing flow for determining the model parameters. The specific content of the set data processing flow will be described in detail in subsequent embodiments.

[0040] Step S302: Use multiple second point cloud data to verify the model parameters of the left wall plane model of the carriage, and extract multiple seventh point cloud data from the multiple second point cloud data according to the first verification results corresponding to the multiple second point cloud data.

[0041] Here, after determining the model parameters of the left wall plane model of the carriage, the point cloud data belonging to the left wall plane of the carriage can be determined from multiple second point cloud data using these parameters. Specifically, the model parameters of the left wall plane model of the carriage are verified using multiple second point cloud data. The left wall plane model of the carriage can be represented by the plane equation A_left*x1+B_left*y1+C_left*z1=D_left. Substituting each second point cloud data into the plane equation of the left wall plane model of the carriage yields the corresponding result error. The result error is the first verification result corresponding to the second point cloud data. The result error refers to the difference between the output result of the second point cloud data based on the plane equation of the left wall plane model of the carriage and D_left. For example, assuming the second point cloud data is (x1, y1, z1), substituting the second point cloud data into the above plane equation yields the corresponding output result D1, where D... 1＝ A_left*x1+B_left*y1+C_left*z1, and then determine the first test result corresponding to the second point cloud data (x1, y1, z1) based on |D1-D_left|.

[0042] After determining the first test result corresponding to the second point cloud data, the first test result is compared with a set threshold. If the first test result is greater than the set threshold, it means that the corresponding second point cloud data is point cloud data outside the left wall of the carriage. Therefore, the corresponding second point cloud data can be identified as the sixth point cloud data, and multiple sixth point cloud data can be extracted from multiple second point cloud data. In practical applications, since multiple sixth point cloud data are point cloud data located outside the left wall of the carriage, meaning that multiple sixth point cloud data contain point cloud data located on the right wall of the carriage, these multiple sixth point cloud data can be processed to determine the model parameters of the right wall planar model of the carriage.

[0043] Step S303: Process multiple seventh-point cloud data according to the set data processing flow to determine the model parameters of the right wall plane model of the carriage.

[0044] In this embodiment, the point cloud data with x > 0 in multiple seventh point cloud data are processed according to the set data processing flow. The point cloud data with x > 0 in multiple seventh point cloud data are used to fit the right wall plane of the carriage, thereby determining the model parameters of the right wall plane model of the carriage. The model parameters of the right wall plane model of the carriage include A_right, B_right, C_right and D_right, where A_right corresponds to the X-axis parameter of the right wall plane model of the carriage, B_right corresponds to the Y-axis parameter of the right wall plane model of the carriage, C_right corresponds to the Z-axis parameter of the right wall plane model of the carriage, and D_right corresponds to the constant term of the right wall plane model of the carriage.

[0045] Step S304: Determine the model parameters of the center plane model of the carriage based on the model parameters of the left wall plane model and the right wall plane model of the carriage.

[0046] Here, the model parameters of the car's center plane model can be determined by the model parameters of the car's left wall plane model and the car's right wall plane model. Specifically, the model parameters of the car's center plane model can be determined by the following relationships:

[0047] In one embodiment, such as Figure 4 As shown, multiple fifth-point cloud data are determined based on multiple second-point cloud data, including:

[0048] Step S401: Use multiple seventh-point cloud data to verify the model parameters of the right wall plane model of the carriage, and extract and determine multiple eighth-point cloud data from the multiple seventh-point cloud data according to the second verification results corresponding to the multiple seventh-point cloud data.

[0049] In this embodiment, the fifth point cloud data refers to the point cloud data located outside the side walls and interior floor of the carriage. Based on this, multiple seventh point cloud data need to be processed to obtain multiple fifth point cloud data. Here, the model parameters of the carriage interior floor model can be determined, and then the point cloud data belonging to the carriage interior floor and the point cloud data located outside the carriage interior floor can be determined using the model parameters of the carriage interior floor model, thereby extracting multiple fifth point cloud data.

[0050] Here, point cloud data belonging to the right wall plane of the carriage is separated from multiple seventh point cloud data, thus obtaining multiple eighth point cloud data. The model parameters of the carriage's interior ground model are obtained by processing these eighth point cloud data. Specifically, the right wall plane model of the carriage can be represented by the plane equation A_right*x + B_right*y + C_right*z = D_right. The model parameters of the right wall plane model are verified using multiple seventh point cloud data, yielding the result error corresponding to each seventh point cloud data, and thus obtaining the second verification result corresponding to the multiple seventh point cloud parameters. Assuming the existence of seventh point cloud data (x2, y2, z2), substituting the seventh point cloud data into the above plane equation yields the output result D2, where the output result D2 = A_right*x2 + B_right*y2 + C_right*z2. The second verification result corresponding to the seventh point cloud data (x2, y2, z2) is then determined based on |D2 - D_right|.

[0051] The second test result is compared with a set threshold. If the second test result is greater than the set threshold, it indicates that the corresponding seventh point cloud data is point cloud data outside the plane of the right wall of the carriage. The corresponding seventh point cloud data is then identified as the eighth point cloud data, and multiple eighth point cloud data can be extracted from multiple seventh point cloud data. It should be noted that since the eighth point cloud data is actually one of the multiple seventh point cloud data, and the seventh point cloud data is point cloud data located outside the plane of the left wall of the carriage, the eighth point cloud data is point cloud data located outside the planes of both the left and right walls of the carriage.

[0052] Step S402: Process multiple eighth-point cloud data according to the set data processing flow to determine the model parameters of the ground model inside the carriage.

[0053] Here, the data processing flow is set as the data processing flow for determining the model parameters of the ground model inside the carriage. By processing multiple eighth-point cloud data according to the set data processing flow, the model parameters of the ground model inside the carriage can be determined. The model parameters of the ground model inside the carriage include A_ground, B_ground, C_ground and D_ground. A_ground corresponds to the X-axis parameter of the ground model inside the carriage, B_ground corresponds to the Y-axis parameter of the ground model inside the carriage, C_ground corresponds to the Z-axis parameter of the ground model inside the carriage, and D_ground corresponds to the constant term of the ground model inside the carriage.

[0054] Step S403: Use multiple eighth-point cloud data to verify the model parameters of the ground model inside the carriage, and extract and determine multiple fifth-point cloud data from the multiple eighth-point cloud data according to the third verification results corresponding to the multiple eighth-point cloud data.

[0055] In this embodiment, after determining the model parameters of the ground model inside the carriage, the ground model inside the carriage can be represented by the planar equation A_ground*x + B_ground*y + C_ground*z = D_ground. The model parameters of the ground model inside the carriage are checked using multiple eighth point cloud data. This checking process can be used to determine whether the eighth point cloud data is point cloud data located on the ground inside the carriage, and the result error corresponding to each eighth point cloud data can be obtained. Then, the third check result corresponding to the eighth point cloud data can be determined. For example, assuming there is an eighth point cloud data (x3, y3, z3), substituting the eighth point cloud data into the above planar equation can obtain the corresponding output result D3, where the output result D3 = A_ground*x3 + B_ground*y3 + C_ground*z3. Then, the third check result corresponding to the eighth point cloud data (x3, y3, z3) is determined according to |D3 - D_ground|.

[0056] The third test result is compared with a set threshold. If the third test result is greater than the set threshold, it indicates that the corresponding eighth point cloud data is located outside the floor of the carriage. The eighth point cloud data with a third test result greater than the set threshold is identified as the fifth point cloud data. Multiple fifth point cloud data are then extracted from these multiple eighth point cloud data. In practical applications, since the eighth point cloud data is extracted from the seventh point cloud data, the multiple eighth point cloud data are essentially a subset of the seventh point cloud data. Furthermore, the seventh point cloud data is located outside the side walls of the carriage. Therefore, the multiple eighth point cloud data are point cloud data located outside both the side walls of the carriage and the floor.

[0057] It should be noted that the threshold values ​​set in steps S302, S401 and S403 can be different values, and the specific threshold values ​​can be set according to actual needs.

[0058] The following section provides a detailed explanation of the data processing workflow, such as... Figure 5 As shown, the data processing flow is configured as follows:

[0059] Step S501: Assign corresponding cells to multiple point cloud data, and use the cell number corresponding to the multiple point cloud data as an index to store the multiple point cloud data into the first array.

[0060] Here, multiple point cloud data are processed into voxel grids, and a corresponding cell is assigned to each point cloud data. The sequence number of the corresponding cell is used as the index of the point cloud data, thereby realizing the storage of multiple point cloud data. Here, the cell is obtained by dividing a set area.

[0061] The point cloud data can be the sixth point cloud data, the seventh point cloud data, or the eighth point cloud data. In this processing step, the implementation details of the processing steps differ depending on the type of point cloud data being processed.

[0062] When the point cloud data is the sixth point cloud data, the YZ plane is used as the grid plane. The sixth point cloud data is then processed into voxel grids and sorted. Specifically, the minimum Y-axis coordinate and the minimum Z-axis coordinate (y_min, z_min) among multiple sixth point cloud data are used as the voxel origins. The maximum Y-axis coordinate (y_max), minimum Y-axis coordinate (y_min), maximum Z-axis coordinate (z_max), and minimum Z-axis coordinate (z_min) are used as the corner points of the grid. The grid boundaries are determined by y_max, y_min, z_max, and z_min. Within this boundary region, the region is divided into different cells according to a set cell size, and the cells are numbered in a row-value sorting manner. Figure 6 As shown, Figure 6 The diagram shows the cells divided on the YZ plane. Multiple sixth-point cloud data points are traversed, and a corresponding cell number is assigned to each sixth-point cloud data point. This assigned cell number serves as an index for the sixth-point cloud data point, allowing each sixth-point cloud data point to be stored in a first array, where the first array, vector1, is a two-dimensional array.

[0063] When the point cloud data is the seventh point cloud data, the YZ plane is used as the grid plane, and multiple seventh point cloud data are processed into voxel meshes. Similarly, the minimum Y-axis coordinate and the minimum Z-axis coordinate (y_min, z_min) of multiple seventh point cloud data are used as the voxel origin. The maximum Y-axis coordinate (y_max), minimum Y-axis coordinate (y_min), maximum Z-axis coordinate (z_max), and minimum Z-axis coordinate (z_min) of multiple seventh point cloud data are used as the corner points of the grid. That is, y_max, y_min, z_max, and z_min can form the grid boundary. The grid area formed by the boundary is divided by setting the cell size, so that the network area formed by the boundary is divided into multiple different cells. These cells are labeled with corresponding numbers according to the row value. All seventh point cloud data are traversed, and a corresponding cell number is assigned to each seventh point cloud data. Using the cell number value of the seventh point cloud data as an index, the seventh point cloud data is stored in the first array, which is a two-dimensional array.

[0064] When the point cloud data is the eighth point cloud data, the XZ plane is used as the grid plane, and multiple eighth point cloud data are processed into voxel meshes. Specifically, the minimum X-axis coordinate and minimum Z-axis coordinate (x_min, z_min) of multiple eighth point cloud data are used as the voxel origin. The maximum X-axis coordinate (x_max), minimum X-axis coordinate (x_min), maximum Z-axis coordinate (z_max), and minimum Z-axis coordinate (z_min) of multiple eighth point cloud data are used as the corner points of the mesh, that is, x_max, x_min, z_max, and z_min can form the boundary of the mesh. The mesh area formed by the boundary is divided by setting the cell size, so that the network area formed by the boundary is divided into multiple different cells, and these cells are labeled with corresponding numbers according to the row value sorting method, such as... Figure 7 As shown, Figure 7 The diagram shows the cells divided on the XZ plane. Multiple eighth-point cloud data points are traversed, and a corresponding cell number is assigned to each eighth-point cloud data point. Using the cell number value as an index, each eighth-point cloud data point is stored in a first array, which is a two-dimensional array.

[0065] Step S502: Based on the location of the cells assigned to the multiple point cloud data, store the neighboring point cloud data of each index into the second array.

[0066] This step is used to determine the neighboring point cloud data for each point cloud data. Since each point cloud data is assigned a corresponding cell, and the cell number serves as the index of the corresponding point cloud data, the neighboring point cloud data for each point cloud data can be determined by referring to the position of each cell on the grid plane. The neighboring point cloud data for each index is then stored in the second array, which is a two-dimensional array.

[0067] The following details the methods for determining the neighborhood data of different point cloud datasets:

[0068] When multiple point cloud data points are considered as multiple sixth point cloud data points, the sixth point cloud data stored in the first array is traversed from top to bottom to determine the neighboring point cloud data for each index. The neighboring point cloud data includes the sixth point cloud data corresponding to the current index, the left index, and the right index in the first array, as well as the point cloud data corresponding to the cell index in the same column of the second array, where the row number minus one is the cell number. For example, refer to... Figure 6 The grid plane shown is searched from top to bottom to find the point cloud data corresponding to index 3 in the second array, which includes the point cloud data corresponding to index 2, 3, and 4 in the first array. The point cloud data corresponding to index 7 in the second array is also searched, which includes the point cloud data corresponding to index 2, 3, and 4 in the first array, as well as the point cloud data corresponding to index 3 in the second array.

[0069] When multiple point cloud data are the seventh point cloud data, the seventh point cloud data stored in the first array is traversed from top to bottom to determine the neighboring point cloud data of each index. The neighboring point cloud data includes the point cloud data corresponding to the current index, its left index, and its right index in the first array, and also includes the point cloud data corresponding to the cell index of row number minus 1 in the same column of the second array.

[0070] When multiple point cloud datasets are considered as multiple eighth point cloud datasets, the second array stores neighboring point cloud data at different indices. The point cloud data in the second array is traversed from top to bottom to find the point cloud data corresponding to the grid plane at the corresponding index in the first array. For example, refer to... Figure 7 The grid plane shown is in Figure 7In the grid plane shown, the point cloud data corresponding to index 5 in the first array is searched. The neighbors of index 5 are indices 2, 4, 5, 6, 8, and 9. Specifically, the neighbors of indices 2 and 4 are indices 1, 2 and 6 are indices 3, 4 and 8 are indices 7, and 6 and 8 are indices 9. Based on this, the neighboring point cloud data of index 5 in array vector2 can be determined to be indices 1, 2, 3, 4, 5, 6, 7, 8, and 9. The method for determining the neighboring point cloud data here is different from the methods used to determine the neighbors of the sixth and seventh point cloud data.

[0071] Step S503: Save the point cloud data corresponding to the indices that meet the set conditions in the second array to the third array.

[0072] Here, the set conditions are mainly used to exclude point cloud data located outside the carriage plane from the point cloud data. Then, the carriage plane model can be constructed using the point cloud data belonging to the carriage plane. The point cloud data belonging to the carriage plane is saved to the third array, which is a one-dimensional array.

[0073] When multiple point cloud datasets are considered as multiple sixth point cloud datasets, the second array stores the neighborhood data of the sixth point cloud dataset. The x_max - x_min of the point cloud data at each index in the second array is calculated, and the result is compared with a first set threshold. If the result is less than the first set threshold, the sixth point cloud data at the corresponding index in the first array is saved to the third array. Here, by comparing the calculation result with the first set threshold, point cloud data outside the plane of the left wall of the carriage can be excluded. For example, if the x_max - x_min of the sixth point cloud data at index 3 in the second array is less than the first set threshold, the sixth point cloud data at the corresponding index in the first array can be saved; that is, the sixth point cloud data corresponding to index 3 in the first array is saved.

[0074] When multiple point cloud datasets are considered as multiple seventh point cloud datasets, the second array stores the neighborhood data of the seventh point cloud dataset. The x_max - x_min of the point cloud data at each index in the second array is calculated, and the result is compared with a first set threshold. If the result is less than the first set threshold, the point cloud data at the corresponding index in the first array is saved to the third array. Here, by comparing the calculation result with the first set threshold, point cloud data outside the plane of the right wall of the carriage can be excluded.

[0075] It should be noted that the first set threshold in processing the seventh point cloud data and the first set threshold in processing the sixth point cloud data can be different values.

[0076] When multiple point cloud datasets are classified as eighth point cloud datasets, the second array stores the neighborhood data of the eighth point cloud dataset. The y_max - y_min of the point cloud data at each index in the second array is calculated, and the result is compared with a second set threshold. If the result is less than the second set threshold, the point cloud data at the corresponding index in the first array is saved to the third array. Here, by comparing the calculation result with the second set threshold, point cloud data outside the ground inside the carriage can be excluded.

[0077] Step S504: Voxel downsampling is performed on the third array to obtain the fourth array.

[0078] Here, the third array is downsampled to obtain the fourth array. Downsampling the third array reduces the amount of data that needs to be processed, thereby improving the processing efficiency of point cloud data.

[0079] Step S505: Use the set algorithm to process the fourth array to determine the initial model parameters of the corresponding carriage plan model.

[0080] Here, the algorithm refers to the Random Sampling Consensus (RANSAC) algorithm. The fourth array is fitted to a plane using the RANSAC algorithm, which specifically includes the following steps:

[0081] Step 1): Randomly select 3 point cloud data points from the multiple point cloud data points in the fourth array, and substitute these 3 point cloud data points into the plane equation A*x+B*y+C*z=D respectively to calculate the initial model parameters A0, B0, C0, and D0 of the carriage planar model. Here, when the multiple point cloud data points are multiple sixth point cloud data points, the calculated initial model parameters are for the left wall planar model of the carriage; when the multiple point cloud data points are multiple seventh point cloud data points, the calculated initial model parameters are for the right wall planar model of the carriage; and when the multiple point cloud data points are multiple eighth point cloud data points, the calculated initial model parameters are for the ground model inside the carriage.

[0082] Step 2) Using other point cloud data from multiple point cloud datasets, substitute them into the plane equation constructed based on the initial model parameters of the carriage plane model determined in Step 1), that is, substitute them into the plane equation A0*x+B0*y+C0*z=D0, and the distance d between different points and the carriage plane can be calculated.

[0083] Step 3) Compare the distance d with the set threshold (usually set to the thickness of the carriage wall). If the distance d is less than the set threshold, the point cloud data is determined to be an interior point. Count the number of interior points under the model parameters corresponding to the plane equation A0*x+B0*y+C0*z=D0, and save the initial model parameters corresponding to the current number of interior points.

[0084] Step 4): Repeat steps 1) to 3). If the current number of interior points is greater than the maximum number of interior points already saved, update the initial model parameters of the carriage plan model based on the initial model parameters corresponding to the current number of interior points. For example, if the saved initial model parameters are A0, B0, C0, and D0, and assuming the current number of interior points is greater than the maximum number of interior points already saved, and the initial model parameters corresponding to the current number of interior points are A1, B1, C1, and D1, then the initial model parameters A1, B1, C1, and D1 can be saved, and the corresponding plane equation can be updated to A1*x + B1*y + C1*z = D1. This step is to ensure that the determined initial model parameters of the carriage plan model are the same as those corresponding to the maximum number of interior points.

[0085] Step 5): When the threshold is reached, update the final initial model parameters.

[0086] Step S506: Use the least squares method to perform plane fitting on the initial model parameters of the carriage plan model to obtain the model parameters of the corresponding carriage plan model.

[0087] Here, the least squares method is used to perform planar fitting on the initial model parameters of the determined carriage planar model, and finally obtain the model parameters of the carriage planar model. Specifically, when multiple point cloud data are the sixth point cloud data, the model parameters A_left, B_left, C_left, and D_left of the carriage left wall planar model can be obtained; when multiple point cloud data are the seventh point cloud data, the model parameters A_right, B_right, C_right, and D_right of the carriage right wall planar model can be obtained; and when multiple point cloud data are the eighth point cloud data, the model parameters A_ground, B_ground, C_ground, and D_ground of the carriage floor model can be obtained.

[0088] It should be noted that, in this embodiment, the multiple seventh-point cloud data used to set the data processing flow refers to the seventh-point cloud data in the portion where x > 0.

[0089] Step S202: Determine the parameters of each axis in the center coordinate system of the carriage based on the model parameters of the carriage center plane model.

[0090] Here, the car center coordinate system is established with the point (0, 0) on the X-axis and Y-axis of the car center plane as the origin. The model parameters (A_middle, B_middle, C_middle) of the car center plane model are used as the X-axis parameters of the car center coordinate system, and (0, 1, 0) is used as the Y-axis parameters. The Z-axis parameters of the car center coordinate system can be obtained through cross product calculation. Specifically, the process of determining the Z-axis parameters is as follows:

[0091]

[0092] Based on this, the X-axis, Y-axis, and Z-axis parameters of the car center coordinate system can be determined.

[0093] Step S203: Based on the axis parameters in the center coordinate system of the carriage, convert multiple fifth point cloud data into multiple corresponding third point cloud data.

[0094] Here, based on the X-axis, Y-axis and Z-axis parameters in the center coordinate system of the carriage, multiple fifth point cloud data can be converted into corresponding point cloud data in the center coordinate system of the carriage, and then the third point cloud data corresponding to each fifth point cloud data can be obtained.

[0095] Step S103: Determine the fourth point cloud data from among multiple third point cloud data.

[0096] Here, data analysis is performed on multiple third-point cloud data to determine the fourth-point cloud data. The fourth-point cloud data represents the data of the target cargo in the laser coordinate system. In practical applications, the target cargo is one of multiple cargoes stored in the vehicle compartment. For example, the target cargo can be the cargo closest to the unmanned forklift.

[0097] It should be noted that the laser coordinate system is the coordinate system of the lidar configured on the unmanned forklift, the third point cloud data is the point cloud data in the center coordinate system of the truck bed, and the fourth point cloud data is the target cargo data in the laser coordinate system. Therefore, there is a conversion between two coordinate systems in the process of determining the fourth point data. The conversion of the point cloud data in the center coordinate system of the truck bed to the point cloud data in the laser coordinate system will be explained in detail in the following content.

[0098] In one embodiment, such as Figure 8 As shown, among multiple third-point cloud data, fourth-point cloud data is determined, including:

[0099] Step S801: Based on the pre-issued target coordinates, divide the multiple third point cloud data into a first group of third point cloud data and a second group of third point cloud data.

[0100] The target coordinates here are the same as those in step S202. In this embodiment, the target coordinates are assumed to be (x0, y0, z0, th0). The main purpose is to divide multiple third point cloud data into a first group and a second group based on the sign of the y-value in the target coordinates. In the first group, each third point cloud data point has x < 0, and in the second group, each point cloud data point has x > 0, where x is in the laser coordinate system. Generally, y0 > 0 corresponds to the point cloud data point where x > 0.

[0101] Step S802: Perform Euclidean clustering on the first group of third point cloud data and the second group of third point cloud data respectively to determine the third point cloud data corresponding to the target goods.

[0102] Here, Euclidean clustering is performed on the first group of third point cloud data and the second group of third point cloud data respectively to determine whether the third point cloud data belongs to the goods. Then, the point cloud data of the target goods is determined from it. Specifically, the following steps are included:

[0103] Step 1) Set the minimum cargo size parameter and generate a detection box based on the minimum cargo size parameter. Using the detection box, multiple third point cloud data with sizes that match the detection box can be obtained from the first group of third point cloud data and the second group of third point cloud data. In other words, the point cloud data that can be initially identified as belonging to cargo can be determined.

[0104] Step 2) Sort the third point cloud data obtained in Step 1) in ascending order in the z direction, and determine the point cloud data corresponding to the smallest z value as the third point cloud data of the target goods. That is, determine the goods closest to the unmanned forklift as the target goods, and obtain the third point cloud data corresponding to the target goods.

[0105] S803 processes the third point cloud data corresponding to the target cargo and converts the third point cloud data corresponding to the target cargo into fourth point cloud data.

[0106] Here, the third point cloud data corresponding to the target cargo is the data in the center coordinate system of the carriage. Processing the third point cloud data corresponding to the target cargo can convert the point cloud data of the target cargo in the center coordinate system of the carriage into point cloud data in the laser coordinate system. Specifically, the third point cloud data of the target cargo is inverted to obtain the fourth point cloud data of the target cargo in the laser coordinate system.

[0107] Step S104: Process the fourth point cloud data to generate the driving path of the unmanned forklift, and control the unmanned forklift to drive according to the driving path.

[0108] Processing the fourth cloud data can generate a driving path for the unmanned forklift to reach the target goods. The driving path can include the driving distance, driving direction, etc. The unmanned forklift is positioned and driven according to the driving path, so that the unmanned forklift can identify the location of the target goods through the driving path.

[0109] In one embodiment, such as Figure 9 As shown, multiple first point cloud data collected by the LiDAR installed on the unmanned forklift inside the vehicle are processed to obtain multiple second point cloud data, including:

[0110] Step S901: Set the first filtering parameters based on the size of the unmanned forklift, and filter multiple first point cloud data according to the first filtering parameters.

[0111] Here, a first data filtering process is performed on multiple first point cloud datasets to remove those belonging to the unmanned forklift. Specifically, a first filtering parameter is set based on the dimensions of the unmanned forklift, including its length and width. The filter processes the multiple first point cloud datasets according to the first filtering parameter, resulting in multiple first point cloud datasets that do not belong to the unmanned forklift.

[0112] Step S902: Correct the multiple first point cloud data obtained after filtering according to the preset target coordinates.

[0113] When the unmanned forklift reaches the monitoring point, the perception algorithm module of the unmanned forklift sends the target coordinates. Assuming the target coordinates are (x0, y0, z0, th0), where x0 is the target coordinate value on the X-axis, y0 is the target coordinate value on the Y-axis, z0 is the target coordinate value on the Z-axis, and th0 is the angle value of the target coordinate. Here, the target coordinates can be understood as a set threshold. Based on the angle value th0 in the target coordinates, the angle of the multiple first point cloud data obtained after filtering is corrected.

[0114] Step S903: Set the second filtering parameters based on the size of the carriage and the third set threshold, and filter the multiple corrected data to be processed according to the second filtering parameters to obtain multiple second point cloud data.

[0115] Here, a second data filtering is performed on the multiple first point cloud data after angle correction. The second filtering parameters are set based on the dimensions of the carriage and a third preset threshold. The dimensions of the carriage include its length, width, and height. The second filtering parameters are set by adding the third preset threshold to the carriage dimensions. The filter uses the second filtering parameters to filter the multiple angle-corrected first point cloud data. The second point cloud data is the retained point cloud data after filtering. For example, the carriage's length is a1, width is b1, and height is c1. The third preset threshold is (a2, b2, c2), and the corresponding second filtering parameters are (a1+a2, b1+b2, c1+c2). That is, point cloud data falling within the (a1+a2, b1+b2, c1+c2) space are determined as second point cloud data. In practical applications, through the second data filtering, point cloud data of interest can be selected for subsequent data processing.

[0116] In one embodiment, such as Figure 10 As shown, based on the data in point four, the driving path of the unmanned forklift is generated, including:

[0117] Step S1001: Determine the angle parameters based on the fourth point cloud data and the model parameters of the carriage center plane model, and project the fourth point cloud data onto the carriage center plane to obtain the projected coordinates.

[0118] Here, the model parameters of the car center plane model include A_middle, B_middle, C_middle and D_middle, respectively. The process for determining the model parameters of the car center plane model can be referred to the content of the above-mentioned related embodiments.

[0119] The fourth point cloud data is the point cloud data in the laser coordinate system. The angle of the Z-axis of the laser coordinate system can be determined in the fourth point cloud data. The angle parameter (yaw) can be determined by the model parameters of the center plane model of the carriage and the angle of the Z-axis of the laser coordinate system.

[0120] The fourth point cloud data is projected onto the central plane of the carriage to obtain the corresponding projection coordinates.

[0121] Step S1002: Generate the driving path based on the angle parameters and projection coordinates.

[0122] Here, the driving path is generated based on the angle parameters and projection coordinates. In other words, the unmanned forklift is positioned and driven according to the angle parameters and projection coordinates, so that the unmanned forklift can drive to the target goods and perform the task of picking up and placing goods according to the driving path.

[0123] This application also provides an application embodiment, see reference. Figure 11 As shown, Figure 11This diagram illustrates the process of determining the travel path of an unmanned forklift.

[0124] Step S1101: Obtain point cloud data for fitting the left / right walls of the carriage.

[0125] Step S1102: Extract point cloud data where x is less than 0, and perform voxel meshing and voxel downsampling processing on the YZ plane.

[0126] Step S1103: Perform plane fitting based on the RANSAC algorithm and iteratively update the model parameters.

[0127] Step S1104: Determine the final model parameters based on the least squares method.

[0128] Step S1105: Determine the model parameters of the ground model inside the carriage.

[0129] Step S1106: Determine the model parameters of the center plane model of the carriage.

[0130] Step S1107: Remove point cloud data from the left and right sides of the carriage walls and the interior floor.

[0131] Step S1108: Using the center plane of the carriage as a reference (y=0), determine whether the point cloud data with y>0 and y<0 are the point cloud data of the cargo.

[0132] Step S1109: Determine the coordinates and angle parameters of the target cargo based on the laser coordinate system.

[0133] Step S1110: Project the coordinates of the target cargo based on the laser coordinate system onto the center plane of the carriage to obtain the coordinates of the projection point.

[0134] Step S1111: The unmanned forklift positions itself and moves according to the projection point coordinates and angle parameters.

[0135] In the above embodiment, multiple point cloud data collected by the lidar configured on the unmanned forklift located inside the compartment are processed to separate the point cloud data belonging to the target goods from the multiple first point cloud data. The driving path of the unmanned forklift is determined based on the point cloud data of the target goods, thereby enabling the unmanned forklift to identify the position of the target goods inside the compartment, avoiding collisions between the unmanned forklift and the compartment walls or goods during movement, and improving the operating efficiency of the unmanned forklift.

[0136] In one embodiment, an unmanned forklift is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement a navigation method for the unmanned forklift.

[0137] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements a navigation method for an unmanned forklift.

[0138] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0139] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0140] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0141] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0142] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A navigation method for an unmanned forklift, characterized in that, include: Multiple first point cloud data collected by a LiDAR system located inside the unmanned forklift are processed to obtain multiple second point cloud data. This process includes: setting a first filtering parameter based on the size of the unmanned forklift and filtering the multiple first point cloud data according to the first filtering parameter; correcting the filtered multiple first point cloud data according to preset target coordinates; setting a second filtering parameter based on the size of the forklift and a third preset threshold, and filtering the corrected first point cloud data according to the second filtering parameter to obtain multiple second point cloud data. The second point cloud data represents the point cloud data belonging to the goods and the point cloud data of the forklift to be processed. The process involves processing multiple second point cloud data sets to extract multiple third point cloud data sets. This includes: determining model parameters of the carriage center plane model and multiple fifth point cloud data sets based on the multiple second point cloud data sets; the fifth point cloud data sets representing point cloud data located outside the carriage side walls and interior floor; determining the axis parameters of each axis in the carriage center coordinate system based on the model parameters of the carriage center plane model; and converting the multiple fifth point cloud data sets into corresponding multiple third point cloud data sets based on the axis parameters of each axis in the carriage center coordinate system; the third point cloud data sets representing point cloud data located inside the carriage in the carriage center coordinate system. Among the multiple third point cloud data, a fourth point cloud data is determined; the fourth point cloud data represents the data of the target cargo in the laser coordinate system; The process of processing the fourth point cloud data to generate the driving path of the unmanned forklift includes: determining angle parameters based on the fourth point cloud data and the model parameters of the center plane model of the truck body, and projecting the fourth point cloud data onto the center plane of the truck body to obtain projection coordinates; generating the driving path based on the angle parameters and the projection coordinates; and controlling the unmanned forklift to travel according to the driving path.

2. The navigation method for an unmanned forklift according to claim 1, characterized in that, The step of determining the model parameters of the carriage center plane model based on multiple second point cloud data includes: Multiple sixth point cloud data are extracted from multiple second point cloud data, and the multiple sixth point cloud data are processed according to the set data processing flow to determine the model parameters of the left wall plane model of the carriage; the sixth point cloud data represents the point cloud data used to fit the left wall of the carriage. The model parameters of the left wall plane model of the carriage are verified using multiple second point cloud data, and multiple seventh point cloud data are extracted from the multiple second point cloud data according to the first verification results corresponding to the multiple second point cloud data; the seventh point cloud data represents the point cloud data located outside the left wall plane of the carriage in the multiple second point cloud data. The data processing flow is set up to process multiple seventh point cloud data to determine the model parameters of the right wall plane model of the carriage. Based on the model parameters of the left wall plane model and the right wall plane model of the carriage, the model parameters of the center plane model of the carriage are determined.

3. The navigation method for an unmanned forklift according to claim 2, characterized in that, Multiple fifth-point cloud data are determined based on multiple second-point cloud data, including: The model parameters of the right wall plane model of the carriage are verified using multiple seventh point cloud data, and multiple eighth point cloud data are extracted and determined from the multiple seventh point cloud data according to the second verification results corresponding to the multiple seventh point cloud data; the eighth point cloud data represents the point cloud data located outside the right wall plane of the carriage in the multiple seventh point cloud data. The data from multiple eighth-point cloud data points are processed according to the established data processing procedure to determine the model parameters of the ground model inside the carriage. The model parameters of the ground model inside the carriage are verified using multiple eighth point cloud data, and multiple fifth point cloud data are extracted and determined from the multiple eighth point cloud data based on the third verification results corresponding to the multiple eighth point cloud data.

4. The navigation method for an unmanned forklift according to claim 3, characterized in that, The defined data processing flow includes the following processing steps: Multiple point cloud data are assigned corresponding cells, and the multiple point cloud data are stored in a first array using the cell numbers as indexes; the cells are obtained by dividing a set area. Based on the location of the cells allocated to the multiple point cloud data, the neighboring point cloud data of each index is stored in the second array; Save the point cloud data corresponding to the indices in the second array that meet the set conditions to the third array; Voxel downsampling is performed on the third array to obtain the fourth array; The fourth array is processed using a set algorithm to determine the initial model parameters of the corresponding carriage plan model; The initial model parameters of the carriage plan model are fitted using the least squares method to obtain the model parameters of the corresponding carriage plan model; wherein, The point cloud data includes the sixth point cloud data, the seventh point cloud data, and the eighth point cloud data; When the point cloud data is the sixth point cloud data or the seventh point cloud data, the setting condition is that the difference between the maximum X-axis value and the minimum X-axis value of the point cloud data in the index of the second array is less than the first set threshold. When the point cloud data is the eighth point cloud data, the setting condition is that the difference between the maximum Y-axis value and the minimum Y-axis value of the point cloud data in the index of the second array is less than the second set threshold.

5. The navigation method for an unmanned forklift according to claim 1, characterized in that, The step of determining the fourth point cloud data from among the multiple third point cloud data includes: Based on the pre-issued target coordinates, the multiple third point cloud data are divided into a first group of third point cloud data and a second group of third point cloud data; Euclidean clustering is performed on the first group of third point cloud data and the second group of third point cloud data respectively to determine the third point cloud data corresponding to the target goods; the third point cloud data corresponding to the target goods is represented as the third point cloud data closest to the unmanned forklift among the third point cloud data belonging to the goods. The third point cloud data corresponding to the target cargo is processed and converted into the fourth point cloud data.

6. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the navigation method of the unmanned forklift as described in any one of claims 1-5.

7. An unmanned forklift, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the navigation method for the unmanned forklift as described in any one of claims 1-5.

Images