Vehicle control method, device and autonomous vehicle

Abstract

The present disclosure provides a vehicle control method and device and an autonomous vehicle, and relates to the technical field of autonomous driving and off-road scene terrain detection. The specific scheme is: obtaining a target point cloud; generating a target grid map according to the position of the vehicle at the time of collecting the target point cloud; performing category recognition on the target point cloud to obtain category confidence of data points in the target point cloud belonging to each preset category; projecting the target point cloud into the target grid map to obtain a corresponding relationship between a grid in the target grid map and a data point in the target point cloud; for each grid in the target grid map, calculating a parameter value of a drivability parameter of the grid according to the position information and the category confidence of the data point corresponding to the grid; and controlling the vehicle to drive according to the target grid map. The embodiment of the present disclosure can improve the smoothness, comfort and safety of the vehicle driving, especially the autonomous driving in the off-road scene.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and particularly to the field of autonomous driving technology, and especially to a vehicle control method, device and autonomous vehicle. Background Technology

[0002] To ensure that autonomous vehicles can drive safely and smoothly, they are typically equipped with sensors such as cameras and lidar to collect environmental information about the vehicle's surroundings. The data processing equipment on board the autonomous vehicle can then detect the surrounding terrain based on the environmental information collected by the sensors, and control the vehicle to drive safely according to the detected terrain.

[0003] For example, in off-road vehicle scenarios, a LiDAR can be installed in the vehicle. The LiDAR can collect point cloud data of the environment around the vehicle. The data processing equipment in the vehicle can detect the terrain of the surrounding off-road environment based on the point cloud data collected by the LiDAR, and then control the vehicle to drive safely based on the detected terrain.

[0004] Because terrain detection can be affected by a variety of environmental factors around the vehicle, such as dynamic obstacles like pedestrians and vehicles, or weather factors like dust, rain, and snow, the accuracy of terrain detection is relatively low, which may lead to lower vehicle control safety. Summary of the Invention

[0005] This disclosure provides a vehicle control method, apparatus, and autonomous vehicle.

[0006] According to one aspect of this disclosure, a vehicle control method is provided, comprising:

[0007] Obtain the target point cloud in real time from the lidar installed on the vehicle;

[0008] A target grid map is generated based on the vehicle's position at the time of acquisition of the target point cloud;

[0009] The target point cloud is classified to obtain the category confidence of the data points in the target point cloud belonging to each preset category;

[0010] The target point cloud is projected onto the target grid map to obtain the correspondence between the grids in the target grid map and the data points in the target point cloud;

[0011] For each grid cell in the target grid map, the drivability parameter of the grid cell is calculated based on the location information of the data point corresponding to the grid cell and the category confidence level. The drivability parameter of the grid cell is used to describe the drivability of the area corresponding to the grid cell in the environment where the vehicle is located.

[0012] The vehicle is controlled to move based on the target grid map after calculating drivability parameters.

[0013] According to another aspect of this disclosure, a vehicle control device is provided, comprising:

[0014] The point cloud acquisition module is used to acquire the target point cloud collected in real time by the lidar installed on the vehicle;

[0015] The map generation module is used to generate a target raster map based on the vehicle's position at the time of acquisition of the target point cloud;

[0016] The category recognition module is used to perform category recognition on the target point cloud and obtain the category confidence of the data points in the target point cloud belonging to each preset category;

[0017] A point cloud projection module is used to project the target point cloud onto the target grid map to obtain the correspondence between the grids in the target grid map and the data points in the target point cloud.

[0018] The parameter calculation module is used to calculate the drivability parameter value of each grid in the target grid map based on the location information of the data point corresponding to the grid and the category confidence level. The drivability parameter of the grid is used to describe the drivability of the area corresponding to the grid in the environment where the vehicle is located.

[0019] The vehicle control module is used to control the vehicle's movement based on the target grid map after calculating drivability parameters.

[0020] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0021] At least one processor; and

[0022] A memory communicatively connected to the at least one processor; wherein,

[0023] The memory stores instructions that can be executed by the at least one processor, which, when executed, enable the at least one processor to perform the vehicle control method described above.

[0024] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform the vehicle control method described above.

[0025] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described vehicle control method.

[0026] According to another aspect of this disclosure, an autonomous vehicle is provided, including the electronic devices as described above.

[0027] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, after obtaining the target point cloud collected by the lidar, the category confidence of the data points in the target point cloud belonging to each preset category is identified. Through projection, the correspondence between grids and data points in the target grid map is obtained. This allows the calculation of the drivability parameters of each grid using the location information and category confidence of the data points corresponding to each grid. Since the drivability parameters of a grid describe the drivability of the area corresponding to that grid in the vehicle's environment, the drivability of the area corresponding to each grid in the surrounding environment can be accurately determined based on the target grid map after calculating the drivability parameters. This enables the vehicle to be controlled to travel within areas with high drivability. Therefore, applying the vehicle control solution provided in this embodiment can improve the smoothness and safety of vehicle driving.

[0028] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0029] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0030] Figure 1 A schematic flowchart illustrating the first vehicle control method provided in this embodiment of the present disclosure;

[0031] Figure 2 A schematic flowchart illustrating a second vehicle control method provided in an embodiment of this disclosure;

[0032] Figure 3 A flowchart illustrating a third vehicle control method provided in this embodiment of the present disclosure;

[0033] Figure 4 A flowchart illustrating the fourth vehicle control method provided in this embodiment of the present disclosure;

[0034] Figure 5 This is a schematic diagram of the structure of a first type of vehicle control device provided in an embodiment of the present disclosure;

[0035] Figure 6 This is a schematic diagram of the structure of a second vehicle control device provided in an embodiment of the present disclosure;

[0036] Figure 7 This is a schematic diagram of the structure of a third vehicle control device provided in an embodiment of the present disclosure;

[0037] Figure 8 This is a schematic diagram of the structure of the fourth vehicle control device provided in the embodiments of this disclosure;

[0038] Figure 9 This is a block diagram of an electronic device used to implement the vehicle control method of the embodiments of this disclosure. Detailed Implementation

[0039] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0040] During vehicle control, the accuracy of terrain detection is low due to environmental factors in the surrounding environment, which may lead to lower vehicle control safety.

[0041] In addition, in off-road scenarios, vehicles lack high-precision maps of the environment. This means that vehicles need to detect the surrounding terrain in real time while driving and control their movement based on the detected terrain. However, the terrain around vehicles in off-road scenarios is usually rugged and uneven. During vehicle movement, the coordinate axes of the LiDAR on the vehicle deviate significantly from the horizontal ground, causing the point cloud obtained by the LiDAR scan to deviate from the actual terrain distribution, thus increasing the difficulty of static terrain detection.

[0042] To address the aforementioned problems, this disclosure provides a vehicle control method and apparatus. Specific embodiments are described below.

[0043] See Figure 1 , Figure 1 This is a flowchart illustrating a first vehicle control method provided in an embodiment of the present disclosure. In this embodiment, the method includes the following steps S101-S106.

[0044] Step S101: Obtain the target point cloud collected in real time by the lidar installed on the vehicle.

[0045] One or more lidar sensors can be installed on the aforementioned vehicles.

[0046] When a single lidar is installed on the aforementioned vehicle, it can be mounted on the front of the vehicle to collect point cloud data of the area in front of the vehicle. When multiple lidars are installed on the aforementioned vehicle, these multiple lidars can be installed at multiple different locations on the vehicle. When obtaining the target point cloud, the point cloud data collected by these multiple lidars at the same or similar times can be obtained as the target point cloud.

[0047] Step S102: Generate a target raster map based on the vehicle's position at the time of target point cloud acquisition.

[0048] The size of the grid cells in the target grid map can be preset by the user. For example, 0.25 meters, 0.2 meters, etc.

[0049] Specifically, when obtaining the target point cloud, the acquisition time of the target point cloud can be obtained. This allows the vehicle's position at the acquisition time of the target point cloud to be determined in the vehicle's historical driving trajectory. Based on the determined position, a target raster map can be generated.

[0050] In one embodiment of this disclosure, when generating a target grid map based on the vehicle's location, the relative position between the area collected by the LiDAR and the vehicle can be obtained in advance. Thus, after determining the vehicle's location, the location of the area collected by the LiDAR at the time of target point cloud acquisition can be determined based on the vehicle's location and the relative position, thereby generating a target grid map at that location.

[0051] In addition, since LiDAR can collect point cloud data of the vehicle's surrounding environment in 360 degrees, a grid map can be generated with the vehicle's position as the center after obtaining the vehicle's position.

[0052] Step S103: Perform category identification on the target point cloud to obtain the category confidence of the data points in the target point cloud belonging to each preset category.

[0053] The aforementioned preset categories can be categories determined based on factors that may occur in the environment.

[0054] For example, the aforementioned preset categories may include ground, dynamic obstacles, noise, and background. The dynamic obstacle category may be the category of objects that can move in the environment, such as pedestrians and vehicles. The noise category may be the category of objects that exist in the environment, such as dust or objects generated by rain, snow, fog, etc. The background category may be considered as the category of other objects besides ground, dynamic obstacles, and noise.

[0055] The aforementioned preset categories may also include positive obstacle categories and negative obstacle categories. Positive obstacle categories include retaining walls, trees, rocks, etc., while negative obstacle categories include potholes, road depressions, cut-off sections, etc.

[0056] Specifically, existing category recognition models and algorithms can be used to classify target point clouds. For example, existing 3D-MiniNet, SqueezeSegV3, and RangeNet network models can be used to classify target point clouds, obtaining the category confidence score of each data point in the identified target point cloud belonging to each preset category. The higher the category confidence score of a data point belonging to a preset category, the greater the probability that the data point belongs to that preset category; the lower the category confidence score of a data point belonging to a preset category, the less likely that the data point belongs to that preset category.

[0057] Step S104: Project the target point cloud onto the target grid map to obtain the correspondence between the grids in the target grid map and the data points in the target point cloud.

[0058] Specifically, the target point cloud mentioned above is the point cloud data of the vehicle's surrounding environment, and the target grid map mentioned above can also be regarded as a grid map of the vehicle's surrounding environment. By projecting the target point cloud onto the target grid map, it is possible to determine the grid to which each data point in the target point cloud belongs in the target grid map after projection. In this way, the correspondence between the data points in the target point cloud and the grids in the target grid map is obtained.

[0059] Projecting the target point cloud onto the target raster map can be achieved using existing projection techniques, which will not be detailed here.

[0060] Step S105: For each grid cell in the target grid map, calculate the parameter value of the drivability parameter of the grid cell based on the location information of the data point corresponding to the grid cell and the category confidence level.

[0061] The drivability parameter of the grid is used to describe the drivability of the area corresponding to the grid in the vehicle's environment.

[0062] The drivability parameter can range from 0 to 1. A drivability parameter of 0 indicates that the area corresponding to the grid is not drivable, while a drivability parameter of 1 indicates that the area corresponding to the grid is a flat and drivable area. A drivability parameter between 0 and 1 indicates that the area corresponding to the grid is an uneven area with potholes.

[0063] For details on how to calculate the feasibility parameter value based on the location information of the data points and the category confidence level, please refer to the following embodiments, which will not be described in detail here.

[0064] Step S106: Control the vehicle's movement based on the target grid map after calculating the drivability parameters.

[0065] Specifically, after calculating the drivability parameters of each grid in the target grid map, the drivability of the area corresponding to each grid can be determined from the drivability parameters of each grid, that is, the drivability of the entire environment corresponding to the target grid map. In this way, based on the target grid map after calculating the drivability parameters, the area with higher drivability in the environment around the vehicle can be identified, thereby controlling the vehicle to drive in the area with higher drivability.

[0066] As can be seen from the above, when controlling a vehicle using the solution provided in this disclosure, after obtaining the target point cloud collected by the lidar, the category confidence of the data points in the target point cloud belonging to each preset category is identified. Through projection, the correspondence between grids and data points in the target grid map is obtained. This allows the calculation of the drivability parameters of each grid using the location information and category confidence of the data points corresponding to each grid. Since the drivability parameters of a grid describe the drivability of the area corresponding to the grid in the vehicle's environment, the drivability of the area corresponding to each grid in the surrounding environment can be accurately determined based on the target grid map after calculating the drivability parameters. This enables the vehicle to be controlled to travel within areas with high drivability. Therefore, the vehicle control solution provided in this disclosure can improve the smoothness and safety of vehicle driving. Furthermore, the solution provided in this disclosure can be applied to various scenarios such as off-road driving and urban traffic, and has a wide range of applications.

[0067] In addition, the vehicle control scheme provided by the embodiments of this disclosure only requires point cloud data collected by lidar as input information, which can reduce the cost of vehicle control.

[0068] The following explains the specific implementation method for calculating the drivability parameters of the grid.

[0069] In one embodiment of this disclosure, see [link to embodiment]. Figure 2 The present invention provides a flowchart of a second vehicle control method. In this embodiment, the above method includes the following steps S201-S208.

[0070] Step S201: Obtain the target point cloud collected in real time by the lidar installed on the vehicle.

[0071] Step S202: Generate a target raster map based on the vehicle's position at the time of target point cloud acquisition.

[0072] Step S203: Perform category identification on the target point cloud to obtain the category confidence of the data points in the target point cloud belonging to each preset category.

[0073] Step S204: Project the target point cloud onto the target raster map to obtain the correspondence between the raster in the target raster map and the data points in the target point cloud.

[0074] The steps S201-S204 are the same as the steps S101-S104 above, and will not be repeated here.

[0075] Step S205: For each grid cell in the target grid map, determine the category of the grid cell based on the category confidence of the data points corresponding to that grid cell.

[0076] In one embodiment of this disclosure, for each grid, the corresponding data points can be determined, and each data point has a category confidence level corresponding to each preset category. Thus, for each category, the mean category confidence level of each data point belonging to that category can be calculated. After calculating the mean category confidence levels for each category, the category corresponding to the maximum mean can be determined as the category of that grid.

[0077] For example, if there are two categories A and B, and the raster corresponds to three data points K1, K2, and K3, the category confidence scores of data point K1 belonging to categories A and B are D1a and D1b, respectively; the category confidence scores of data point K2 belonging to categories A and B are D2a and D2b, respectively; and the category confidence scores of data point K3 belonging to categories A and B are D3a and D3b, respectively, then for category A, the mean category confidence score can be calculated as (D1a + D2a + D3a) / 3, and for category B, the mean category confidence score can be calculated as (D1b + D2b + D3b) / 3.

[0078] In this scheme, the category of the raster is determined by comprehensively considering the category confidence of each data point corresponding to the raster. The information considered is more comprehensive. Thus, the category of the raster can be accurately determined based on the calculated average category confidence of each category.

[0079] In another embodiment of this disclosure, for each data point, after obtaining the category confidence of the data point belonging to each preset category, the category corresponding to the maximum category confidence of the data point can be determined as the category of the data point. Thus, for each grid, after determining the data point corresponding to the grid, the number of data points of each category in the grid can be counted, thereby determining the category corresponding to the maximum number as the category of the grid.

[0080] In this scheme, the category corresponding to the maximum category confidence of each data point is determined as the category of each data point. By counting the number of data points of each category in the raster, the category corresponding to the maximum number is found, which can accurately and quickly determine the category of the raster.

[0081] Step S206: If the grid belongs to a preset dynamic object category, then the drivability parameter of the grid is determined to be a preset value.

[0082] The aforementioned preset values ​​can be parameter values ​​located at half, one-third, or other positions within the range of feasible parameters, or they can be other values ​​set by the user based on their own experience.

[0083] The aforementioned dynamic object categories can be categories pre-specified by the user from the preset categories. For example, after setting four categories—ground, dynamic obstacles, noise, and background—dynamic obstacles and noise can be specified as dynamic object categories.

[0084] Specifically, after determining the category of the grid, it can be determined whether the category of the grid belongs to the category of dynamic objects. If it is, a preset value can be directly given as the parameter value of the grid's feasibility parameter. If it is not, it can be calculated through the following S207.

[0085] Step S207: If the category of the grid does not belong to the dynamic object category, then determine the target data point in the data points corresponding to the grid, calculate the feature values ​​of various preset terrain features of the grid according to the location information of the target data point, and perform weighted calculation on each calculated feature value based on the preset weight corresponding to each preset terrain feature to obtain the weighted calculation result, which is used as the parameter value of the drivability parameter of the grid.

[0086] The terrain features of the grid are used to characterize the terrain of the corresponding area in the vehicle's environment.

[0087] The aforementioned preset weights can be set manually.

[0088] In one embodiment of this disclosure, the target data point can be determined by any one of the following three implementation methods.

[0089] In the first implementation, a data point that is close to the surrounding data points can be identified among the data points corresponding to the grid and used as the target data point.

[0090] In the second implementation, the data points corresponding to the maximum class confidence that do not belong to the category of dynamic objects can be identified as target data points among the data points corresponding to the grid.

[0091] In the third implementation, when obtaining the target point cloud, the laser reflectivity of each data point in the target point cloud can also be obtained. Thus, when determining the target data points, the category confidence level of each data point and the laser reflectivity are combined.

[0092] For example, data points whose laser reflectivity is greater than a preset reflectivity and whose category corresponding to the maximum category confidence does not belong to the above dynamic object category can be identified as target data points.

[0093] The preset reflectivity can be set manually.

[0094] In this scheme, the target data point is determined by using laser reflectivity and the confidence level of the data point belonging to each category. The information used is relatively large, which can accurately determine the target data point and thus improve the accuracy of subsequent processing.

[0095] In one embodiment of this disclosure, when determining the target data point corresponding to a grid, in addition to determining it from the data points corresponding to the grid itself, the target data point can also be determined from the data points corresponding to neighboring grids.

[0096] The neighboring grid includes the grid itself and other grids adjacent to it.

[0097] For example, the aforementioned neighborhood grid may include the grid itself and four other grids that are adjacent to it vertically, horizontally, or to the left and right; the aforementioned neighborhood grid may also include the grid itself and eight grids surrounding it.

[0098] Specifically, for each grid cell, the neighboring grid cells can be determined, and the data points corresponding to each grid cell in the neighboring grid cells can be determined. Thus, the target data point corresponding to the grid cell can be determined from the data points corresponding to each grid cell.

[0099] In this scheme, the target data point can be determined from the data points corresponding to the neighboring grids of the grid. Since there are more data points corresponding to the neighboring grids, determining the target data point from the data points corresponding to the neighboring grids can increase the number of target data points. In this way, the parameter values ​​of the drivability parameters of the grid can be calculated based on the location information of more target data points, which can improve the accuracy of the calculated parameters and thus improve the accuracy of vehicle control.

[0100] After determining the target data points, feature values ​​of various preset terrain features of the grid can be calculated based on the location information of the target data points. Then, based on the preset weights corresponding to each preset terrain feature, the calculated feature values ​​are weighted to obtain the weighted calculation result, which is used as the parameter value of the drivability parameter of the grid.

[0101] In one embodiment of this disclosure, the terrain feature includes at least one of the following features:

[0102] Maximum height difference, maximum height difference above ground, slope, and flatness characteristics.

[0103] Among them, the flatness feature is used to characterize the flatness of the ground corresponding to the grid.

[0104] Taking the maximum height difference as an example, after determining the target data points mentioned above, we can determine the maximum height and minimum height among the heights of each target data point. Subtracting the maximum height from the minimum height will give us the maximum height difference.

[0105] The calculation methods for the characteristic values ​​of maximum ground clearance, slope, and flatness can be found in subsequent embodiments, and will not be detailed here.

[0106] In this scheme, when calculating the feature values ​​of various terrain features, the feature values ​​of the maximum height difference, the maximum height difference above the ground, the slope, and the flatness feature can be calculated. Each of these four terrain features can accurately characterize the terrain of the area corresponding to the grid. Therefore, by calculating multiple features among these four features, the parameter values ​​of the grid's feasibility parameters can be accurately calculated.

[0107] Because this solution considers not only the height of the data point corresponding to the grid in the point cloud space, but also the height of the data point above the ground and the slope of the area corresponding to the grid when calculating the drivability parameters of the grid, it can accurately detect the drivability parameters of rugged areas in off-road scenarios. Based on the target grid map after calculating the drivability parameters, it can accurately control the vehicle to drive safely and smoothly in off-road scenarios.

[0108] Step S208: Control the vehicle's movement based on the target grid map after calculating the drivability parameters.

[0109] This step is the same as step S106 above, and will not be repeated here.

[0110] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, if the grid type is a dynamic object type, the parameter value of the drivability parameter of the grid is determined to be a preset value. If the grid type is not a dynamic object type, the parameter value of the drivability parameter of the grid is calculated by calculating the feature values ​​of various terrain features of the grid. This can eliminate the interference of dynamic objects on the detected terrain, thereby improving the anti-interference ability of vehicle control and the robustness of terrain drivability determination.

[0111] In one embodiment of this disclosure, when the terrain features include the maximum height difference above the ground, the feature value of the terrain features of the grid is calculated as the maximum height difference above the ground of the grid.

[0112] In this case, after classifying the target point cloud and obtaining the class confidence of the data points in the target point cloud belonging to each preset class, the ground data points in the target point cloud that correspond to the class with the highest class confidence can be determined as the ground. Based on the location information of the ground data points in the target point cloud, the plane where the ground is located can be fitted in the space where the target point cloud is located.

[0113] For each grid cell, the distance between each target data point corresponding to the grid cell and the plane on the ground can be calculated. The difference between the maximum and minimum distances among the calculated distances of each target data point is then determined as the maximum height difference above the ground.

[0114] Fitting the plane of the ground based on the location information of ground data points can be achieved using existing ground fitting techniques, which will not be detailed here.

[0115] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, by fitting the ground plane into the space where the target point cloud is located, the maximum ground clearance difference of each grid can be calculated based on the ground plane. If the maximum ground clearance difference of a grid is large, it indicates that there is an obstacle in the area corresponding to that grid, thus the drivability of the area corresponding to that grid is low. It is evident that by using the maximum ground clearance difference, the drivability parameter value of the grid can be accurately calculated, thereby improving the accuracy of vehicle control based on more accurate drivability parameters.

[0116] In one embodiment of this disclosure, when the terrain features include slope, calculating the feature value of the terrain features of the grid is equivalent to calculating the slope corresponding to the grid.

[0117] In this case, based on the location information of each target data point corresponding to the raster, a plane containing each target data point can be fitted in the space where the target point cloud is located. This plane can be understood as the ground of the area corresponding to the raster. After fitting the plane containing each target data point, the normal vector of the plane can be determined, and the angle between the normal vector and the preset direction can be determined as the slope.

[0118] The aforementioned preset direction can be the vertical direction, horizontal direction, etc., of the space where the target point cloud is located.

[0119] Fitting the plane containing each target data point in the space of the target point cloud can be achieved using existing plane fitting techniques, which will not be detailed here.

[0120] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, the plane where each target data point in the grid is located is essentially the ground of the area corresponding to the grid. Therefore, by determining the angle between the normal vector of the fitted plane and the preset direction, the slope characteristics of the grid can be accurately obtained. By utilizing the more accurate slope characteristics, the drivability parameters of the grid can be accurately calculated, thereby improving the accuracy of vehicle control based on the more accurate drivability parameters.

[0121] In one embodiment of this disclosure, when the terrain features include flatness features, the feature value of the terrain features of the grid is calculated as the feature value of the flatness features of the grid.

[0122] In this case, the height standard deviation of each target data point can be calculated as the feature value of the flatness characteristic.

[0123] Specifically, after determining multiple target data points of the raster, the height of each target data point can be determined. Then, the standard deviation formula can be used to calculate the height standard deviation of each target data point, which serves as the feature value of the flatness feature.

[0124] As can be seen from the above, when controlling a vehicle using the scheme provided in this embodiment, since the standard deviation can reflect the dispersion of the data, the height standard deviation of each target data point is calculated. This height standard deviation reflects the dispersion of the height of each target data point. The smaller the height standard deviation, the more concentrated the heights of the target data points are, the smaller the differences are, and the flatter the area corresponding to the grid is. Therefore, by calculating the height standard deviation of each target data point, the characteristic value of the flatness feature can be accurately obtained. Based on the more accurate flatness feature, the parameter value of the grid's drivability parameter can be accurately calculated, thereby improving the accuracy of vehicle control based on the more accurate drivability parameter.

[0125] When there are few target data points corresponding to a grid, the accuracy of calculating slope and flatness features based on the location information of the target data points is low.

[0126] Therefore, after determining the target data points corresponding to the grid, the number of target data points can be counted. If the number of targets is greater than or equal to a preset number, it indicates that there are many target data points corresponding to the grid. In this case, based on the location information of the target data points, the feature values ​​of four characteristics of the grid—maximum height difference, maximum ground clearance difference, slope, and flatness—can be calculated. The feature values ​​of these four characteristics are then weighted to obtain a weighted calculation result, which serves as the parameter value for the drivability parameter of the grid. If the number of targets is less than the preset number, it indicates that there are few target data points corresponding to the grid. In this case, based on the location information of the target data points, only the maximum height difference and maximum ground clearance difference of the grid can be calculated. The feature values ​​of these two characteristics are then weighted to obtain a weighted calculation result, which serves as the parameter value for the drivability parameter of the grid.

[0127] The preset quantity can be a manually set quantity, such as 3 or 4.

[0128] The implementation methods for calculating the feature values ​​of various terrain features can be found in the aforementioned embodiments, and will not be repeated here.

[0129] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, if there are many target data points corresponding to the grid, a large number of feature values ​​of terrain features can be calculated. Based on these multiple terrain features, the parameter values ​​of drivability parameters can be accurately calculated. If there are few target data points corresponding to the grid, slope and flatness features with low accuracy can be omitted from the calculation. Only the maximum height difference and maximum ground clearance difference can be calculated. This avoids the influence of slope and flatness features on the parameter values ​​of drivability parameters, thereby improving the accuracy of the calculated drivability parameter values. Consequently, the accuracy of vehicle control can be improved based on more accurate drivability parameters.

[0130] When projecting a target point cloud onto a target raster map, due to the sparsity of the target point cloud, there may be raster maps where the number of data points corresponding to a raster is zero, making it difficult to calculate the drivability parameter value of that raster.

[0131] To address the aforementioned issues, in one embodiment of this disclosure, see [link to relevant documentation]. Figure 3 The present invention provides a flowchart of a third vehicle control method. In this embodiment, the above method includes the following steps S301-S308.

[0132] Step S301: Obtain the target point cloud collected in real time by the lidar installed on the vehicle.

[0133] Step S302: Generate a target raster map based on the vehicle's position at the time of target point cloud acquisition.

[0134] Step S303: Perform category identification on the target point cloud to obtain the category confidence of the data points in the target point cloud belonging to each preset category.

[0135] Step S304: Project the target point cloud onto the target raster map to obtain the correspondence between the raster in the target raster map and the data points in the target point cloud.

[0136] Step S305: For each grid cell in the target grid map, calculate the parameter value of the drivability parameter of the grid cell based on the location information of the data point corresponding to the grid cell and the category confidence level.

[0137] The steps S301-S305 are the same as the steps S101-S105 above, and will not be repeated here.

[0138] Step S306: If there are unknown grids in the target grid map whose drivability parameters have not been calculated, then predict the drivability parameter value of the unknown grid based on the drivability parameters of the surrounding grids that have been calculated.

[0139] In one embodiment of this disclosure, the parameter value of the drivability parameter of an unknown grid can be predicted based on Bayesian generalized kernel inference, according to the calculated drivability parameters of the grids surrounding the unknown grid.

[0140] Specifically, when predicting the drivability parameters of unknown lattices based on Bayesian generalized kernel inference, the drivability parameters of unknown lattices can be used as Bayesian regression, and the drivability parameters of unknown lattices can be modeled as a Gaussian distribution. Its prior distribution is Use the drivability parameters of the known grid as the observation. Predict the drivability parameters for unknown grid cells.

[0141] The distribution of drivability parameters for an unknown grid can be represented by the following expression:

[0142]

[0143] Where, μ * x represents the mean drivability of an unknown grid cell. * This represents the spatial coordinates of an unknown raster, where i represents the i-th known raster, N represents the total number of known raster cells, and x represents the unknown raster cell. i Represents the spatial coordinates of the i-th known grid cell, y i Let represent the drivability parameter of the i-th known grid cell, k represent the kernel function, μ represent the mean drivability parameter of the unknown grid cell, μ0 represent the prior mean drivability parameter of the unknown grid cell, and σ represent the drivability parameter of the unknown grid cell. 2 The variance of the dability parameters of the unknown grid is represented by λ, and the prior influence coefficient is represented by d. i This represents the spatial distance between the unknown grid cell and the i-th known grid cell.

[0144] The mean and variance of the feasibility parameters for the unknown raster are:

[0145]

[0146]

[0147] in,

[0148] After calculating the mean value of the feasibility parameters of an unknown raster, this mean value can be used as the feasibility parameter of the unknown raster.

[0149] In this scheme, based on the calculated drivability parameters of the surrounding grates, and using Bayesian generalized kernel inference, the drivability parameter values ​​of the unknown grates can be accurately determined. Therefore, based on the more accurate drivability parameter values, the accuracy of subsequent processing can be improved.

[0150] In another embodiment of this disclosure, the weight of the feasibility parameter of a known grid can be determined based on the relative position between the known grid and the unknown grid and the preset correspondence. The feasibility parameters of each known grid are then weighted and calculated to obtain the weighted calculation result, which is used as the parameter value of the feasibility parameter of the unknown grid.

[0151] Among them, the known grid is the grid around the unknown grid whose feasibility parameters have been calculated.

[0152] The preset correspondence is: the relationship between relative position and weight that the user pre-sets based on their own experience.

[0153] Specifically, we can identify the known grids surrounding the unknown grid and obtain their relative positions. Based on these relative positions and the aforementioned correspondence, we can determine the weights of each known grid, perform a weighted calculation on the drivability parameters of each known grid, and obtain the weighted calculation result as the parameter value for the drivability parameters of the unknown grid.

[0154] In this scheme, based on the relative positions and preset correspondences between known and unknown grids, the weights of the drivability parameters of known grids can be accurately determined. Thus, the drivability parameters of each known grid are weighted and calculated to obtain the weighted calculation result, which is the accurate parameter value of the drivability parameters of the unknown grid.

[0155] Step S307: Control the vehicle's movement based on the target grid map after predicting drivability parameters.

[0156] Step 307 is the same as step S106 above, and will not be repeated here.

[0157] As can be seen from the above, when controlling a vehicle using the solution provided in the embodiments of this disclosure, the parameter values ​​of the drivability parameters of unknown grid cells in the target grid map can be predicted. This can ensure that each grid cell in the target grid map is a grid cell with known drivability parameters as much as possible, thereby controlling the vehicle to drive according to the target grid map and improving the reliability of vehicle control.

[0158] The accuracy of single-frame point cloud data acquired by lidar may be low. Therefore, the accuracy of the drivability parameters of each grid in the target grid map based solely on real-time acquired target point cloud computing is low.

[0159] To address the aforementioned issues, in one embodiment of this disclosure, see [link to relevant documentation]. Figure 4 The present invention provides a flowchart of a fourth vehicle control method. In this embodiment, the above method includes the following steps S401-S411.

[0160] Step S401: Obtain the target point cloud collected in real time by the lidar installed on the vehicle.

[0161] Step S402: Generate a target raster map based on the vehicle's position at the time of target point cloud acquisition.

[0162] Step S403: Perform category identification on the target point cloud to obtain the category confidence of the data points in the target point cloud belonging to each preset category.

[0163] Step S404: Project the target point cloud onto the target raster map to obtain the correspondence between the raster in the target raster map and the data points in the target point cloud.

[0164] Step S405: For each grid cell in the target grid map, calculate the parameter value of the drivability parameter of the grid cell based on the location information of the data point corresponding to the grid cell and the category confidence level.

[0165] The steps S401-S405 are the same as the steps S101-S105 above, and will not be repeated here.

[0166] Step S406: Obtain the historical raster map generated before the acquisition time of the target point cloud.

[0167] As the vehicle moves, the LiDAR continuously collects point cloud data of the surrounding environment. At each point cloud acquisition moment, the vehicle's data processor generates a grid map and calculates the drivability parameters of each grid cell. Therefore, the drivability parameters of each grid cell are known in historical grid maps generated before the target point cloud acquisition moment.

[0168] Step S407: Determine the first overlapping raster in the historical raster map corresponding to the same area and the second overlapping raster in the target raster map.

[0169] Specifically, the region corresponding to the historical grid map in the vehicle's surrounding environment can be determined, and the region corresponding to the target grid map in the vehicle's surrounding environment can be determined. The overlapping region of these two regions is detected. The grid corresponding to this overlapping region in the historical grid map is the first overlapping grid, and the grid corresponding to this overlapping region in the target grid map is the second overlapping grid.

[0170] Step S408: Based on the drivability parameters of the first overlapping grid, predict the candidate drivability parameters of the second overlapping grid.

[0171] Specifically, based on the drivability parameters of the first overlapping grid, existing prediction techniques can be applied to predict the candidate drivability parameters of the second overlapping grid.

[0172] Step S409: If the second overlapping grid has drivability parameters, then determine the final drivability parameters of the second overlapping grid based on the drivability parameters of the second overlapping grid and the candidate drivability parameters.

[0173] Specifically, the final drivability parameters of the second overlapping grid can be determined using any of the following three methods.

[0174] In the first implementation, the drivability parameters of the second overlapping grid and the weights of the candidate drivability parameters can be preset. After obtaining the drivability parameters and candidate drivability parameters of the second overlapping grid, the two parameters are weighted and calculated to obtain the final drivability parameters of the second overlapping grid.

[0175] In the second implementation, the average of the drivability parameter and the candidate drivability parameter of the second overlapping grid can be directly taken as the final drivability parameter of the second overlapping grid.

[0176] In the third implementation, Kalman filtering can be used to fuse the drivability parameters of the first overlapping grid and the second overlapping grid.

[0177] When performing Kalman filtering, the drivability parameters of the first overlapping grid can be used as prior information, and the drivability parameters of the second overlapping grid can be used as measurement information. The final drivability parameters of the second overlapping grid are calculated according to the following expression:

[0178]

[0179]

[0180]

[0181]

[0182]

[0183] Among them, a t c represents the model parameters of the prediction model in Kalman filtering. t z represents the model parameters of the measurement model in Kalman filtering. t-1 This represents the drivability parameter of the first overlapping grid at time t-1, which corresponds to the time in the historical grid map mentioned above. ε represents the predicted drivability parameter value of the second overlapping grid at the acquisition time of the target point cloud calculated according to the prediction model. t v represents the process noise of the prediction model. t-1 This represents the variance of the drivability parameter at time t-1. ξ represents the prediction variance of the drivability parameter. t k represents the noise of the measurement model. t Indicates Kalman gain, This represents the measured value of the drivability parameter, specifically the drivability parameter of the second overlapping grid, z. t v represents the posterior estimate of the drivability parameter. t This represents the posterior variance of the drivability parameter.

[0184] Step S410: If the second overlapping grid does not have a feasibility parameter, then the predicted candidate feasibility parameter is determined as the feasibility parameter of the second overlapping grid.

[0185] Step S411: Control the vehicle's movement based on the target grid map after calculating the drivability parameters.

[0186] Step S411 is the same as step S106 above, and will not be repeated here.

[0187] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, the candidate drivability parameters of the second overlapping grid are predicted based on the drivability parameters of the first overlapping grid. Thus, if drivability parameters exist in the second overlapping grid, the drivability parameters of the second overlapping grid are re-determined based on the drivability parameters and candidate drivability parameters, thereby improving the accuracy of the drivability parameters of the second overlapping grid. If no drivability parameters exist in the second overlapping grid, the candidate drivability parameters can be directly determined as the drivability parameters of the second overlapping grid. This ensures that the drivability parameters of each grid in the target grid map are known as much as possible, thereby improving the safety of vehicle control.

[0188] Corresponding to the above-described vehicle control method, this disclosure also provides a vehicle control device.

[0189] In one embodiment of this disclosure, see [link to embodiment]. Figure 5 A schematic diagram of the structure of a first type of vehicle control device is provided. In this embodiment, the device includes:

[0190] The point cloud acquisition module 501 is used to acquire the target point cloud collected in real time by the lidar installed on the vehicle.

[0191] The map generation module 502 is used to generate a target raster map based on the position of the vehicle at the time of acquisition of the target point cloud;

[0192] The category recognition module 503 is used to perform category recognition on the target point cloud and obtain the category confidence of the data points in the target point cloud belonging to each preset category;

[0193] The point cloud projection module 504 is used to project the target point cloud onto the target grid map to obtain the correspondence between the grid in the target grid map and the data points in the target point cloud.

[0194] The parameter calculation module 505 is used to calculate the drivability parameter value of each grid in the target grid map based on the location information of the data point corresponding to the grid and the category confidence level. The drivability parameter of the grid is used to describe the drivability of the area corresponding to the grid in the environment where the vehicle is located.

[0195] The vehicle control module 506 is used to control the vehicle's movement based on the target grid map after calculating drivability parameters.

[0196] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, after obtaining the target point cloud collected by the lidar, the category confidence of the data points in the target point cloud belonging to each preset category is identified. Through projection, the correspondence between grids and data points in the target grid map is obtained. This allows the calculation of the drivability parameters of each grid using the location information and category confidence of the data points corresponding to each grid. Since the drivability parameters of a grid describe the drivability of the area corresponding to that grid in the vehicle's environment, the drivability of the area corresponding to each grid in the surrounding environment can be accurately determined based on the target grid map after calculating the drivability parameters. This enables the vehicle to be controlled to travel within areas with high drivability. Therefore, applying the vehicle control solution provided in this embodiment can improve the smoothness and safety of vehicle driving.

[0197] In one embodiment of this disclosure, see [link to embodiment]. Figure 6 A schematic diagram of a second type of vehicle control device is provided. In this embodiment, the device includes:

[0198] The point cloud acquisition module 601 is used to acquire the target point cloud collected in real time by the lidar installed on the vehicle.

[0199] The map generation module 602 is used to generate a target raster map based on the position of the vehicle at the time of acquisition of the target point cloud;

[0200] The category recognition module 603 is used to perform category recognition on the target point cloud and obtain the category confidence of the data points in the target point cloud belonging to each preset category;

[0201] The point cloud projection module 604 is used to project the target point cloud onto the target grid map to obtain the correspondence between the grid in the target grid map and the data points in the target point cloud.

[0202] The category determination submodule 605 is used to determine the category of each grid in the target grid map based on the category confidence of the data points corresponding to that grid.

[0203] The first determining submodule 606 is used to determine the drivability parameter value of each grid in the target grid map as a preset value if the grid's category belongs to the preset dynamic object category.

[0204] The second determining submodule 607 is used to determine the target data point in the data points corresponding to each grid in the target grid map if the category of the grid does not belong to the dynamic object category. Based on the location information of the target data point, it calculates the feature values ​​of multiple preset terrain features of the grid. Based on the preset weights corresponding to each preset terrain feature, it performs a weighted calculation on each calculated feature value to obtain a weighted calculation result, which is used as the parameter value of the drivability parameter of the grid. The terrain features of the grid are used to characterize the terrain of the area corresponding to the grid in the environment where the vehicle is located.

[0205] The vehicle control module 608 is used to control the vehicle's movement based on the target grid map after calculating drivability parameters.

[0206] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, if the grid type is a dynamic object type, the parameter value of the drivability parameter of the grid is determined to be a preset value. If the grid type is not a dynamic object type, the parameter value of the drivability parameter of the grid is calculated by calculating the feature values ​​of various terrain features of the grid. This can eliminate the interference of dynamic objects on the detected terrain, thereby improving the anti-interference capability of vehicle control.

[0207] In one embodiment of this disclosure, the terrain feature includes at least one of the following features:

[0208] Maximum height difference, maximum height difference above ground, slope, and flatness characteristics.

[0209] In this scheme, when calculating the feature values ​​of various terrain features, the feature values ​​of the maximum height difference, the maximum height difference above the ground, the slope, and the flatness feature can be calculated. Each of these four terrain features can accurately characterize the terrain of the area corresponding to the grid. Therefore, by calculating multiple features among these four features, the parameter values ​​of the grid's feasibility parameters can be accurately calculated.

[0210] In one embodiment of this disclosure, when the terrain feature includes the maximum height difference above ground, the device further includes:

[0211] The ground fitting module 609 is used to fit the plane where the ground is located in the space where the target point cloud is located based on the location information of the ground data points in the target point cloud after class identification of the target point cloud and obtaining the class confidence of the data points in the target point cloud belonging to each preset class. The ground data points are: data points whose class is the ground corresponding to the maximum class confidence.

[0212] The second determining submodule 607 is specifically used for:

[0213] Calculate the distance between each target data point and the plane containing the ground;

[0214] The difference between the maximum and minimum distances among the calculated distances to each target data point is determined as the maximum ground clearance difference.

[0215] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, by fitting the ground plane into the space where the target point cloud is located, the maximum ground clearance difference of each grid can be calculated based on the ground plane. If the maximum ground clearance difference of a grid is large, it indicates that there is an obstacle in the area corresponding to that grid, thus the drivability of the area corresponding to that grid is low. It is evident that by using the maximum ground clearance difference, the drivability parameter value of the grid can be accurately calculated, thereby improving the accuracy of vehicle control based on more accurate drivability parameters.

[0216] In one embodiment of this disclosure, when the terrain feature includes the slope, the second determining submodule 607 is specifically used for:

[0217] Based on the location information of each target data point, fit the plane where each target data point is located in the space of the target point cloud;

[0218] The angle between the normal vector of the fitted plane and the preset direction is determined as the slope.

[0219] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, the plane where each target data point in the grid is located is essentially the ground of the area corresponding to the grid. Therefore, by determining the angle between the normal vector of the fitted plane and the preset direction, the slope characteristics of the grid can be accurately obtained. By utilizing the more accurate slope characteristics, the drivability parameters of the grid can be accurately calculated, thereby improving the accuracy of vehicle control based on the more accurate drivability parameters.

[0220] In one embodiment of this disclosure, when the terrain feature includes the flatness feature, the second determining submodule 607 is specifically used for:

[0221] Calculate the height standard deviation of each target data point, and use it as the feature value of the flatness feature.

[0222] As can be seen from the above, when controlling a vehicle using the scheme provided in this embodiment, since the standard deviation can reflect the dispersion of the data, the height standard deviation of each target data point is calculated. This height standard deviation reflects the dispersion of the height of each target data point. The smaller the height standard deviation, the more concentrated the heights of the target data points are, the smaller the differences are, and the flatter the area corresponding to the grid is. Therefore, by calculating the height standard deviation of each target data point, the characteristic value of the flatness feature can be accurately obtained. Based on the more accurate flatness feature, the parameter value of the grid's drivability parameter can be accurately calculated, thereby improving the accuracy of vehicle control based on the more accurate drivability parameter.

[0223] In one embodiment of this disclosure, the second determining submodule 607 is specifically used for:

[0224] Count the number of targets for the target data points;

[0225] If the number of targets is greater than or equal to the preset number, then based on the location information of the target data points, the feature values ​​of the maximum height difference, maximum height difference from the ground, slope, and flatness characteristics of the grid are calculated.

[0226] If the number of targets is less than the preset number, then the maximum height difference and the maximum height difference above the ground of the grid are calculated based on the location information of the target data points.

[0227] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, if there are many target data points corresponding to the grid, a large number of feature values ​​of terrain features can be calculated. Based on these multiple terrain features, the parameter values ​​of drivability parameters can be accurately calculated. If there are few target data points corresponding to the grid, slope and flatness features with low accuracy can be omitted from the calculation. Only the maximum height difference and maximum ground clearance difference can be calculated. This avoids the influence of slope and flatness features on the parameter values ​​of drivability parameters, thereby improving the accuracy of the calculated drivability parameter values. Consequently, the accuracy of vehicle control can be improved based on more accurate drivability parameters.

[0228] In one embodiment of this disclosure, the category determination submodule 605 is specifically used for:

[0229] For each category, calculate the mean category confidence score of each data point corresponding to that grid that belongs to that category;

[0230] The category corresponding to the maximum mean is determined as the category of the raster.

[0231] In this scheme, the category of the raster is determined by comprehensively considering the category confidence of each data point corresponding to the raster. The information considered is more comprehensive. Thus, the category of the raster can be accurately determined based on the calculated average category confidence of each category.

[0232] In one embodiment of this disclosure, the category determination submodule 605 is specifically used for:

[0233] For each data point corresponding to the raster, the category corresponding to the maximum category confidence of the data point is determined as the category of the data point;

[0234] Count the number of data points of each category in the raster;

[0235] The category corresponding to the maximum quantity is determined as the category of the raster.

[0236] In this scheme, the category corresponding to the maximum category confidence of each data point is determined as the category of each data point. By counting the number of data points of each category in the raster, the category corresponding to the maximum number is found, which can accurately and quickly determine the category of the raster.

[0237] In one embodiment of this disclosure, the second determining submodule 607 is specifically used for:

[0238] Among the data points corresponding to the grid, the data points whose laser reflectivity is greater than the preset reflectivity and whose category corresponding to the maximum category confidence does not belong to the category of the dynamic object are identified as target data points.

[0239] In this scheme, the target data point is determined by using laser reflectivity and the confidence level of the data point belonging to each category. The information used is relatively large, which can accurately determine the target data point and thus improve the accuracy of subsequent processing.

[0240] In one embodiment of this disclosure, the second determining submodule 607 is specifically used for:

[0241] The target data point is determined from the data points corresponding to the neighboring rasters of the grid, wherein the neighboring rasters include the grid and other rasters adjacent to the grid.

[0242] In this scheme, the target data point can be determined from the data points corresponding to the neighboring grids of the grid. Since there are more data points corresponding to the neighboring grids, determining the target data point from the data points corresponding to the neighboring grids can increase the number of target data points. In this way, the parameter values ​​of the drivability parameters of the grid can be calculated based on the location information of more target data points, which can improve the accuracy of the calculated parameters and thus improve the accuracy of vehicle control.

[0243] In one embodiment of this disclosure, see [link to embodiment]. Figure 7 A schematic diagram of a third type of vehicle control device is provided. In this embodiment, the device includes:

[0244] The point cloud acquisition module 701 is used to acquire the target point cloud collected in real time by the lidar installed on the vehicle.

[0245] The map generation module 702 is used to generate a target raster map based on the position of the vehicle at the time of acquisition of the target point cloud;

[0246] The category recognition module 703 is used to perform category recognition on the target point cloud and obtain the category confidence of the data points in the target point cloud belonging to each preset category;

[0247] The point cloud projection module 704 is used to project the target point cloud onto the target grid map to obtain the correspondence between the grid in the target grid map and the data points in the target point cloud.

[0248] The parameter calculation module 705 is used to calculate the drivability parameter value of each grid in the target grid map based on the location information of the data point corresponding to the grid and the category confidence level. The drivability parameter of the grid is used to describe the drivability of the area corresponding to the grid in the environment where the vehicle is located.

[0249] The parameter prediction submodule 706 is used to predict the drivability parameter value of the unknown grid cell based on the drivability parameters of the surrounding grid cells if there is an unknown grid cell in the target grid cell map for which drivability parameters have not been calculated.

[0250] The vehicle control submodule 707 is used to control the vehicle's movement based on the target grid map after predicting drivability parameters.

[0251] As can be seen from the above, when controlling a vehicle using the solution provided in the embodiments of this disclosure, the parameter values ​​of the drivability parameters of unknown grid cells in the target grid map can be predicted. This can ensure that each grid cell in the target grid map is a grid cell with known drivability parameters as much as possible, thereby controlling the vehicle to drive according to the target grid map and improving the reliability of vehicle control.

[0252] In one embodiment of this disclosure, the parameter prediction submodule 706 is specifically used for:

[0253] Based on the calculated drivability parameters of the surrounding grates, the parameter values ​​of the drivability parameters of the unknown grates are predicted using Bayesian generalized kernel inference.

[0254] In this scheme, based on the calculated drivability parameters of the surrounding grates, and using Bayesian generalized kernel inference, the drivability parameter values ​​of the unknown grates can be accurately determined. Therefore, based on the more accurate drivability parameter values, the accuracy of subsequent processing can be improved.

[0255] In one embodiment of this disclosure, the parameter prediction submodule 706 is specifically used for:

[0256] The weight of the feasibility parameter of the known grid is determined based on the relative position between the known grid and the unknown grid and a preset correspondence. The known grid is the grid around the unknown grid whose feasibility parameters have been calculated. The preset correspondence is the relationship between the relative position and the weight.

[0257] The feasibility parameters of each known grid are weighted and calculated to obtain the weighted calculation result, which is used as the parameter value of the feasibility parameter of the unknown grid.

[0258] In this scheme, based on the relative positions and preset correspondences between known and unknown grids, the weights of the drivability parameters of known grids can be accurately determined. Thus, the drivability parameters of each known grid are weighted and calculated to obtain the weighted calculation result, which is the accurate parameter value of the drivability parameters of the unknown grid.

[0259] In one embodiment of this disclosure, see [link to embodiment]. Figure 8 A schematic diagram of a fourth type of vehicle control device is provided. In this embodiment, the device includes:

[0260] The point cloud acquisition module 801 is used to acquire the target point cloud collected in real time by the lidar installed on the vehicle.

[0261] The map generation module 802 is used to generate a target raster map based on the position of the vehicle at the time of acquisition of the target point cloud;

[0262] The category recognition module 803 is used to perform category recognition on the target point cloud and obtain the category confidence of the data points in the target point cloud belonging to each preset category;

[0263] The point cloud projection module 804 is used to project the target point cloud onto the target grid map to obtain the correspondence between the grid in the target grid map and the data points in the target point cloud.

[0264] The parameter calculation module 805 is used to calculate the drivability parameter value of each grid in the target grid map based on the location information of the data point corresponding to the grid and the category confidence level. The drivability parameter of the grid is used to describe the drivability of the area corresponding to the grid in the environment where the vehicle is located.

[0265] Map acquisition module 806 is used to acquire a historical raster map generated before the acquisition time of the target point cloud;

[0266] The grid determination module 807 is used to determine the first overlapping grid in the historical grid map and the second overlapping grid in the target grid map that correspond to the same area.

[0267] The parameter prediction module 808 is used to predict candidate drivability parameters of the second overlapping grid based on the drivability parameters of the first overlapping grid.

[0268] The first determining module 809 is used to determine the final drivability parameter of the second overlapping grid based on the drivability parameter of the second overlapping grid and the candidate drivability parameter if the second overlapping grid has drivability parameters.

[0269] The second determining module 810 is used to determine the predicted candidate feasibility parameter as the feasibility parameter of the second overlapping grid if the second overlapping grid does not have a feasibility parameter.

[0270] The vehicle control module 811 is used to control the vehicle's movement based on the target grid map after calculating drivability parameters.

[0271] As can be seen from the above, when controlling a vehicle using the solution provided in this embodiment, the candidate drivability parameters of the second overlapping grid are predicted based on the drivability parameters of the first overlapping grid. Thus, if drivability parameters exist in the second overlapping grid, the drivability parameters of the second overlapping grid are re-determined based on the drivability parameters and candidate drivability parameters, thereby improving the accuracy of the drivability parameters of the second overlapping grid. If no drivability parameters exist in the second overlapping grid, the candidate drivability parameters can be directly determined as the drivability parameters of the second overlapping grid. This ensures that the drivability parameters of each grid in the target grid map are known as much as possible, thereby improving the safety of vehicle control.

[0272] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0273] In one embodiment of this disclosure, an electronic device is provided, comprising:

[0274] At least one processor; and

[0275] A memory communicatively connected to the at least one processor; wherein,

[0276] The memory stores instructions that can be executed by the at least one processor, which, when executed, enables the at least one processor to perform any of the vehicle control methods described in the foregoing method embodiments.

[0277] In one embodiment of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to perform any of the vehicle control methods described in the foregoing method embodiments.

[0278] In one embodiment of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements any of the vehicle control methods described in the foregoing method embodiments.

[0279] In one embodiment of this disclosure, an autonomous vehicle is provided, including electronic devices as mentioned in the above-described electronic device embodiments.

[0280] Figure 9 A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0281] like Figure 9 As shown, device 900 includes a computing unit 901, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 902 or a computer program loaded from storage unit 908 into random access memory (RAM) 903. RAM 903 may also store various programs and data required for the operation of device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via bus 904. Input / output (I / O) interface 905 is also connected to bus 904.

[0282] Multiple components in device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of monitors, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0283] The computing unit 901 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as vehicle control methods. For example, in some embodiments, the vehicle control method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the vehicle control method described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to perform the vehicle control method by any other suitable means (e.g., by means of firmware).

[0284] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0285] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0286] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0287] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0288] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0289] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0290] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0291] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A vehicle control method, comprising: Obtain the target point cloud in real time from the lidar installed on the vehicle; A target grid map is generated based on the vehicle's position at the time of acquisition of the target point cloud; The target point cloud is classified to obtain the category confidence of the data points in the target point cloud belonging to each preset category; The target point cloud is projected onto the target grid map to obtain the correspondence between the grids in the target grid map and the data points in the target point cloud; For each grid in the target grid map, the category of the grid is determined based on the category confidence of the data point corresponding to the grid. If the category of the grid belongs to the preset dynamic object category, the parameter value of the drivability parameter of the grid is determined to be a preset value. If the category of the grid does not belong to the dynamic object category, the target data point in the data point corresponding to the grid is determined. Based on the location information of the target data point, the feature values ​​of multiple preset terrain features of the grid are calculated. Based on the preset weights corresponding to each preset terrain feature, the calculated feature values ​​are weighted to obtain the weighted calculation result, which is used as the parameter value of the drivability parameter of the grid. The terrain features of the grid are used to characterize the terrain of the area corresponding to the grid in the environment where the vehicle is located. The drivability parameter of the grid is used to describe the drivability of the area corresponding to the grid in the environment where the vehicle is located. The vehicle is controlled to move based on the target grid map after calculating drivability parameters.

2. The method according to claim 1, wherein, The terrain features include at least one of the following features: Maximum height difference, maximum height difference above ground, slope, and flatness characteristics.

3. The method according to claim 2, wherein, When the terrain features include the maximum height difference above the ground, after classifying the target point cloud to obtain the class confidence scores of the data points in the target point cloud belonging to each preset class, the method further includes: Based on the location information of the ground data points in the target point cloud, a plane containing the ground is fitted in the space where the target point cloud is located, wherein the ground data points are data points whose category corresponds to the maximum category confidence. The step of calculating feature values ​​of various preset terrain features of the raster based on the location information of the target data points includes: Calculate the distance between each target data point and the plane containing the ground; The difference between the maximum and minimum distances among the calculated distances to each target data point is determined as the maximum ground clearance difference.

4. The method according to claim 2, wherein, When the terrain feature includes the slope, the step of calculating feature values ​​of various preset terrain features of the grid based on the location information of the target data point includes: Based on the location information of each target data point, fit the plane where each target data point is located in the space of the target point cloud; The angle between the normal vector of the fitted plane and the preset direction is determined as the slope.

5. The method according to claim 2, wherein, When the terrain feature includes the flatness feature, the step of calculating the feature values ​​of various preset terrain features of the raster based on the location information of the target data point includes: Calculate the height standard deviation of each target data point, and use it as the feature value of the flatness feature.

6. The method according to claim 2, wherein, The step of calculating feature values ​​of various preset terrain features of the raster based on the location information of the target data points includes: Count the number of targets for the target data points; If the number of targets is greater than or equal to the preset number, then based on the location information of the target data points, the feature values ​​of the maximum height difference, maximum height difference from the ground, slope, and flatness characteristics of the grid are calculated. If the number of targets is less than the preset number, then the maximum height difference and the maximum height difference above the ground of the grid are calculated based on the location information of the target data points.

7. The method according to any one of claims 1-6, wherein, Determining the category of a raster based on the category confidence score of the data points corresponding to that raster includes: For each category, calculate the mean category confidence score of each data point corresponding to that grid that belongs to that category; The category corresponding to the maximum mean is determined as the category of the raster.

8. The method according to any one of claims 1-6, wherein, Determining the category of a raster based on the category confidence score of the data points corresponding to that raster includes: For each data point corresponding to the raster, the category corresponding to the maximum category confidence of the data point is determined as the category of the data point; Count the number of data points of each category in the raster; The category corresponding to the maximum quantity is determined as the category of the raster.

9. The method according to any one of claims 1-6, wherein, Determining the target data point among the data points corresponding to the raster includes: Among the data points corresponding to the grid, the data points whose laser reflectivity is greater than the preset reflectivity and whose category corresponding to the maximum category confidence does not belong to the category of the dynamic object are identified as target data points.

10. The method according to any one of claims 1-6, wherein, Determining the target data point among the data points corresponding to the raster includes: The target data point is determined from the data points corresponding to the neighboring rasters of the grid, wherein the neighboring rasters include the grid and other rasters adjacent to the grid.

11. The method according to any one of claims 1-6, wherein, The step of controlling vehicle movement based on the target grid map after calculating drivability parameters includes: If there are unknown grids in the target grid map whose drivability parameters have not been calculated, then the drivability parameter value of the unknown grid is predicted based on the drivability parameters of the grids surrounding the unknown grid. The vehicle is controlled to move based on the target grid map after predicting drivability parameters.

12. The method according to claim 11, wherein, The step of predicting the drivability parameter value of the unknown grid based on the calculated drivability parameters of the grids surrounding the unknown grid includes: Based on the calculated drivability parameters of the surrounding grates, the parameter values ​​of the drivability parameters of the unknown grates are predicted using Bayesian generalized kernel inference.

13. The method according to claim 11, wherein, The step of predicting the drivability parameter value of the unknown grid based on the calculated drivability parameters of the grids surrounding the unknown grid includes: The weight of the feasibility parameter of the known grid is determined based on the relative position between the known grid and the unknown grid and a preset correspondence. The known grid is the grid around the unknown grid whose feasibility parameters have been calculated. The preset correspondence is the relationship between the relative position and the weight. The feasibility parameters of each known grid are weighted and calculated to obtain the weighted calculation result, which is used as the parameter value of the feasibility parameter of the unknown grid.

14. The method according to any one of claims 1-6, further comprising: Obtain a historical raster map generated before the acquisition time of the target point cloud; Identify the first overlapping grid in the historical grid map and the second overlapping grid in the target grid map that correspond to the same area; Based on the drivability parameters of the first overlapping grid, predict the candidate drivability parameters of the second overlapping grid; If the second overlapping grid has drivability parameters, then the final drivability parameters of the second overlapping grid are determined based on the drivability parameters of the second overlapping grid and the candidate drivability parameters. If the second overlapping grid does not have a feasibility parameter, then the predicted candidate feasibility parameter is determined as the feasibility parameter of the second overlapping grid.

15. A vehicle control device, comprising: The point cloud acquisition module is used to acquire the target point cloud in real time collected by the lidar installed on the vehicle; The map generation module is used to generate a target raster map based on the vehicle's position at the time of acquisition of the target point cloud; The category recognition module is used to perform category recognition on the target point cloud and obtain the category confidence of the data points in the target point cloud belonging to each preset category; A point cloud projection module is used to project the target point cloud onto the target grid map to obtain the correspondence between the grids in the target grid map and the data points in the target point cloud. The category determination submodule is used to determine the category of each grid in the target grid map based on the category confidence of the data points corresponding to that grid. The first determining submodule is used to determine the drivability parameter of each grid in the target grid map as a preset value if the grid belongs to the preset dynamic object category. The drivability parameter of the grid is used to describe the drivability of the area corresponding to the grid in the environment where the vehicle is located. The second determining submodule is used to determine the target data point in the data points corresponding to each grid in the target grid map if the category of the grid does not belong to the dynamic object category. Based on the location information of the target data point, the module calculates the feature values ​​of various preset terrain features of the grid. Based on the preset weights corresponding to each preset terrain feature, the module performs a weighted calculation on each calculated feature value to obtain a weighted calculation result, which is used as the parameter value of the drivability parameter of the grid. The terrain features of the grid are used to characterize the terrain of the area corresponding to the grid in the environment where the vehicle is located. The vehicle control module is used to control the vehicle's movement based on the target grid map after calculating drivability parameters.

16. The apparatus according to claim 15, wherein, The terrain features include at least one of the following features: Maximum height difference, maximum height difference above ground, slope, and flatness characteristics.

17. The apparatus according to claim 16, wherein, If the terrain feature includes the maximum ground elevation difference, it also includes: The ground fitting module is used to fit the plane where the ground is located in the space where the target point cloud is located based on the location information of the ground data points in the target point cloud after class identification of the target point cloud and obtaining the class confidence of the data points in the target point cloud belonging to each preset class. The ground data points are: data points whose class is the ground corresponding to the maximum class confidence. The second determining submodule is specifically used for: Calculate the distance between each target data point and the plane containing the ground; The difference between the maximum and minimum distances among the calculated distances to each target data point is determined as the maximum ground clearance difference.

18. The apparatus according to claim 16, wherein, In the case that the terrain feature includes the slope, the second determining submodule is specifically used for: Based on the location information of each target data point, fit the plane where each target data point is located in the space of the target point cloud; The angle between the normal vector of the fitted plane and the preset direction is determined as the slope.

19. The apparatus according to claim 16, wherein, When the terrain feature includes the flatness feature, the second determining submodule is specifically used for: Calculate the height standard deviation of each target data point, and use it as the feature value of the flatness feature.

20. The apparatus according to claim 16, wherein, The second determining submodule is specifically used for: Count the number of targets for the target data points; If the number of targets is greater than or equal to the preset number, then based on the location information of the target data points, the feature values ​​of the maximum height difference, maximum height difference from the ground, slope, and flatness characteristics of the grid are calculated. If the number of targets is less than the preset number, then the maximum height difference and the maximum height difference above the ground of the grid are calculated based on the location information of the target data points.

21. The apparatus according to any one of claims 15-20, wherein, The category determination submodule is specifically used for: For each category, calculate the mean category confidence score of each data point corresponding to that grid that belongs to that category; The category corresponding to the maximum mean is determined as the category of the raster.

22. The apparatus according to any one of claims 15-20, wherein, The category determination submodule is specifically used for: For each data point corresponding to the raster, the category corresponding to the maximum category confidence of the data point is determined as the category of the data point; Count the number of data points of each category in the raster; The category corresponding to the maximum quantity is determined as the category of the raster.

23. The apparatus according to any one of claims 15-20, wherein, The second determining submodule is specifically used for: Among the data points corresponding to the grid, the data points whose laser reflectivity is greater than the preset reflectivity and whose category corresponding to the maximum category confidence does not belong to the category of the dynamic object are identified as target data points.

24. The apparatus according to any one of claims 15-20, wherein, The second determining submodule is specifically used for: The target data point is determined from the data points corresponding to the neighboring rasters of the grid, wherein the neighboring rasters include the grid and other rasters adjacent to the grid.

25. The apparatus according to any one of claims 15-20, wherein, The vehicle control module includes: The parameter prediction submodule is used to predict the drivability parameter value of the unknown grid cell if there is an unknown grid cell in the target grid cell map for which drivability parameters have not been calculated, based on the drivability parameters of the surrounding grid cells. The vehicle control submodule is used to control vehicle movement based on the target grid map after predicting drivability parameters.

26. The apparatus according to claim 25, wherein, The parameter prediction submodule is specifically used for: Based on the calculated drivability parameters of the surrounding grates, the parameter values ​​of the drivability parameters of the unknown grates are predicted using Bayesian generalized kernel inference.

27. The apparatus according to claim 25, wherein, The parameter prediction submodule is specifically used for: The weight of the feasibility parameter of the known grid is determined based on the relative position between the known grid and the unknown grid and a preset correspondence. The known grid is the grid around the unknown grid whose feasibility parameters have been calculated. The preset correspondence is the relationship between the relative position and the weight. The feasibility parameters of each known grid are weighted and calculated to obtain the weighted calculation result, which is used as the parameter value of the feasibility parameter of the unknown grid.

28. The apparatus according to any one of claims 15-20, further comprising: The map acquisition module is used to acquire historical raster maps generated before the acquisition time of the target point cloud; The grid determination module is used to determine the first overlapping grid in the historical grid map and the second overlapping grid in the target grid map that correspond to the same area. The parameter prediction module is used to predict candidate drivability parameters of the second overlapping grid based on the drivability parameters of the first overlapping grid. The first determining module is used to determine the final drivability parameter of the second overlapping grid based on the drivability parameter of the second overlapping grid and the candidate drivability parameter if the second overlapping grid has drivability parameters. The second determining module is used to determine the predicted candidate feasibility parameter as the feasibility parameter of the second overlapping grid if the second overlapping grid does not have a feasibility parameter.

29. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-14.

30. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-14.

31. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-14.

32. An autonomous vehicle, including the electronic equipment as described in claim 29.

Images