SLAM (Simultaneous Localization and Mapping) rapid initial positioning method and system based on high-precision map

Abstract

The invention provides an SLAM (Simultaneous Localization and Mapping) rapid initial positioning method and system based on a high-precision map, and the method comprises the steps: obtaining the pose of a current position of a mobile carrier, and calculating a feature value of the current position; the feature values comprise parameters capable of reflecting features of markers within a certain range around the current position; extracting all feature codes within a certain range around the current position through the pose of the current position; the feature codes are pre-stored, comprise pose information and feature values, and correspond to the laser point cloud clusters; traversing all the feature codes, and searching feature codes consistent with the feature value of the current position, so as to position a corresponding laser point cloud cluster; and completing initialization of the laser positioning system by adopting the corresponding laser point cloud cluster. By adopting the method, large-range search is not needed during large-range laser positioning, and the search efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the field of high-precision positioning, and in particular relates to a SLAM fast initial positioning method and system based on high-precision maps. Background Art

[0002] In order to reduce the delay of laser positioning, the current application of laser SLAM positioning in mass production often first has a laser point cloud feature map, and then performs positioning through real-time laser feature matching of the point cloud feature map. However, since laser SLAM positioning is mainly initialized by fixing the initial position and initializing by reading several fixed initialization points, this positioning method is strongly dependent on the initial point and cannot be initialized at any position. As the range increases, the laser point cloud feature map will become very large. For a large-scale point cloud feature map, the search time is long and the matching accuracy is low.

[0003] How to ensure the rapid selection of the corresponding laser feature point cloud for matching in a large scene is a major difficulty. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a SLAM fast initial positioning method and system based on high-precision maps, which can improve the search efficiency during large-scale laser positioning.

[0005] The technical solution adopted by the present invention to solve the above technical problems is: a SLAM fast initial positioning method based on high-precision map, which is suitable for a mobile carrier equipped with a combined navigation, high-precision map system and laser positioning system.

[0006] Obtaining the position and posture of the current position of the mobile carrier, and calculating the characteristic value of the current position; the characteristic value includes parameters that can reflect the characteristics of the landmarks within a certain range around the current position;

[0007] Extract all feature codes within a certain range around the current position through the posture of the current position; the feature codes are pre-stored, contain posture information, feature values, and correspond to the laser point cloud cluster;

[0008] Traversing all the feature codes, searching for a feature code that is consistent with the feature value of the current position, so as to locate the corresponding laser point cloud cluster;

[0009] The corresponding laser point cloud cluster is used to complete the initialization of the laser positioning system.

[0010] According to the above scheme, the feature code is obtained in the following manner: when establishing a laser point cloud feature map, the initial position and posture of the mobile carrier obtained by the combined navigation is recorded at the initial moment, and the established laser point cloud data is converted to the global coordinate system through the initial position and posture; in the process of establishing the laser point cloud feature map, the high-precision map system is continuously subscribed to through the combined navigation to obtain environmental features within a certain range around the current position; the feature values ​​in the environmental features are extracted, and the feature codes are formed with the posture of the current position, corresponding to the laser point cloud cluster centered on the current position.

[0011] According to the above scheme, the feature code is: a combination of the coordinates of the current position, the attitude angle and the feature value in a certain order.

[0012] According to the above solution, the environmental features include semantic information of lane lines and signboards.

[0013] According to the above scheme, the characteristic values ​​include the total number of lane lines, the average lane line color, the average lane line type, the total number of signboards and the average signboard type.

[0014] According to the above scheme, the characteristic value of the current position is calculated by the high-precision map system.

[0015] According to the above scheme, the certain range around the current position is within 8-20m around the current position.

[0016] According to the above solution, the mobile carrier is a vehicle.

[0017] As a second aspect of the present invention, the present invention also provides a SLAM fast initial positioning system based on a high-precision map, comprising:

[0018] The current position feature value calculation module is used to obtain the position and posture of the current position from the integrated navigation system carried by the mobile carrier and calculate the feature value of the current position; the feature value includes parameters that can reflect the characteristics of landmarks within a certain range around the current position;

[0019] A feature code extraction module is used to extract all feature codes within a certain range around the current position through the posture of the current position; the feature code is pre-stored, includes posture information, feature values, and corresponds to the laser point cloud cluster;

[0020] A laser point cloud positioning module is used to traverse all the feature codes and find a feature code that is consistent with the feature value of the current position, so as to locate the corresponding laser point cloud cluster;

[0021] The initialization positioning module is used to complete the initialization of the laser positioning system using the corresponding laser point cloud cluster.

[0022] As a third aspect of the present invention, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the high-precision map-based SLAM fast initial positioning method is implemented.

[0023] The beneficial effects of the present invention are as follows: by completing feature coding in the existing laser point cloud data in combination with the feature values ​​of the high-precision map, when performing initial positioning again, a rapid verification can be performed based on the feature coding to determine the initial position; when using this method, when performing large-scale laser positioning, there is no need for a large-scale search, and only the feature coding within a certain range around the current position needs to be verified, thereby improving the search efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a pre-completed workflow diagram of an embodiment of the present invention.

[0025] Figure 2 It is a flow chart of a method for formal positioning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0027] Glossary:

[0028] SLAM: Simultaneous Localization and Mapping, real-time positioning and map construction, is mainly used to enable the mobile carrier to position itself according to the location and map during movement, and to build an incremental map based on its own positioning to achieve autonomous positioning and navigation of the mobile carrier.

[0029] High-precision map: A high-precision map used for autonomous driving. As prior data for autonomous driving, it expresses map elements such as roads, lanes, roadside traffic signs and ground markings. The map accuracy can be down to the centimeter level.

[0030] Combined navigation: a navigation system that combines two or more navigation devices.

[0031] Laser positioning system: A system equipped with a laser radar device that matches the real-time laser point cloud with the existing point cloud map to obtain the current position.

[0032] Posture: It represents the position and attitude of an object relative to the reference coordinate system, including the three-dimensional coordinates of the position and the attitude angles. The attitude angles include roll angle, pitch angle and heading angle.

[0033] The present invention provides a SLAM fast initial positioning method based on high-precision maps, which is applicable to a mobile carrier equipped with a combined navigation system (or a single inertial navigation system), a high-precision map system and a laser positioning system. The mobile carrier can be various types of cars, or other vehicles, robots, etc.

[0034] First, before formal positioning, Figure 1 As shown, first perform the following preset work:

[0035] S01. When establishing a laser point cloud feature map, the position and posture of the current position obtained by the combined navigation is recorded at the initialization time, recorded as T1, and the established laser point cloud data is converted to the global coordinate system through T1. Among them, the combined navigation is to obtain the position and posture of the current position, which can also be completed by a single inertial navigation system.

[0036] The laser point cloud data refers to the built laser point cloud map. The original laser point cloud map is in the odometer coordinate system. It needs to be converted to the world coordinate system through the combined navigation pose at the initialization time. The laser point cloud map is used for matching and positioning in the laser positioning system, and is also the collection of "laser point cloud clusters" that need to be searched later.

[0037] S02. In the process of establishing the laser point cloud feature map, the high-precision map system is continuously subscribed through combined navigation to obtain the semantic information of the environmental features around the current location, including lane lines, signboards, etc. For example, the type of signboard is "speed limit sign", and the value is "60km / h". The lane line is "yellow line, solid line".

[0038] S03, extracting environmental features in a certain range around the current position and calculating feature values. The certain range around the current position is determined according to the positioning error of the combined navigation or the actual situation, usually 8-20m, and 10m is taken in this embodiment.

[0039] The characteristic values ​​include: the total number of lane lines, the average lane line color, the average lane line type, the total number of signboards, and the average signboard type. The total number of lane lines is the total number of lane lines in the semantic information of the high-precision map. The average lane line color is obtained by dividing the sum of the color values ​​of each lane line by the total number of lane lines. For example, by parsing the color value of each lane line in the semantic information, adding these color values ​​and dividing them by the total number of lane lines, the lane line color average is obtained. The lane line type average is obtained by dividing the sum of each lane line type by the total number of lane lines. For example, firstly, each lane line type is represented by a number, and the type numbers of each lane line are added and divided by the total number of lane lines to obtain the lane line type average. The total number of signboards is the total number of signboards in the semantic information of the high-precision map. The signboard type average is obtained by dividing the sum of each signboard type by the total number of signboards. For example, firstly, each signboard type is represented by a number, and the type numbers of each signboard are added and divided by the total number of signboards to obtain the signboard type average.

[0040] S04, assigning the above-mentioned characteristic values ​​to the laser mapping system, and encoding and storing them in a certain order in combination with the posture. For example, in one embodiment of the present invention, the storage is performed in the following form:

[0041] <x,y,z,roll,pitch,yaw,Ln,Lc,Lt,Sn,St>

[0042] Among them, x, y, z are the x-axis, y-axis, and z-axis coordinate values ​​in the world coordinate system of the current position, respectively; roll, pitch, and yaw are the roll angle, pitch angle, and heading angle in the attitude angle of the current position, respectively; Ln, Lc, Lt, Sn, and St are the total number of lane lines, the mean lane line color, the mean lane line type, the total number of signboards, and the mean signboard type in the eigenvalues, respectively.

[0043] Each point in the laser point cloud data has a specific x, y, and z coordinates. Therefore, a center point can be determined for each cluster of point clouds. This center point is consistent with the feature code.<x,y,z,roll,pitch,yaw,Ln,Lc,Lt,Sn,St> In the figure, x, y, and z correspond to each other, and the point cloud cluster can be located by x, y, and z.

[0044] Among them, the laser point cloud feature map adds feature coding on the basis of the laser point cloud map. The laser mapping system uses laser radar to build a laser point cloud map of the environment. By assigning the above feature values ​​to the laser mapping system, a laser point cloud feature map is established, that is, a laser point cloud map with feature coding.

[0045] Through the above presets, feature encoding is completed in the existing laser point cloud data in combination with the feature values ​​of the high-precision map. Figure 2As shown, when performing initial positioning, follow the steps below to perform positioning:

[0046] S1. Obtain the posture of the current position of the mobile carrier and calculate the characteristic value of the current position. The posture of the current position of the carrier is obtained from the combined navigation carried by the mobile carrier. The posture of the current position obtained at this time includes the coordinates and attitude angle of the position. The coordinates of the position are not accurate enough and need to be verified. The characteristic value of the current position here includes parameters that can reflect the characteristics of landmarks within a certain range around the current position, which are calculated by the high-precision map system. Markers include: lane lines, signboards, street lights, etc. Marker characteristics include: quantity, color, specific content, etc.

[0047] S2. Extract all feature codes within a certain range around the current position through the position and posture of the current position. The certain range around the current position here is the same as the certain range mentioned above. In large-scale laser positioning, there is no need for a large-scale search, and only the feature code verification within a certain range around the current position is required.

[0048] S3. Traverse all the feature codes to find a feature code that is consistent with the feature value of the current position, so as to locate the corresponding laser point cloud cluster.

[0049] S4, using the corresponding laser point cloud cluster to complete the initialization of the laser positioning system, and then continue to complete the subsequent laser positioning.

[0050] As a second aspect of the present invention, the present invention also provides a SLAM fast initial positioning system based on a high-precision map, comprising:

[0051] The current position feature value calculation module is used to obtain the position and posture of the current position from the integrated navigation system carried by the mobile carrier and calculate the feature value of the current position; the feature value includes parameters that can reflect the characteristics of markers within a certain range around the current position. Markers include lane lines, signboards, street lights, etc. Marker features include quantity, color, specific content, etc.

[0052] The feature code extraction module is used to extract all feature codes within a certain range around the current position through the posture of the current position; the feature code is pre-stored, including posture information, feature values, and corresponds to the laser point cloud cluster. The feature code is obtained in advance, which has been explained above.

[0053] The laser point cloud positioning module is used to traverse all the feature codes and find the feature code that is consistent with the feature value of the current position, so as to locate the corresponding laser point cloud cluster.

[0054] The initialization positioning module is used to complete the initialization of the laser positioning system using the corresponding laser point cloud cluster.

[0055] As a third aspect of the present invention, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the high-precision map-based SLAM fast initial positioning method is implemented.

[0056] The present invention actually utilizes the idea of ​​storing and retrieving laser feature point cloud maps. In the existing laser point cloud data, the feature values ​​of the high-precision map are combined to complete the feature coding for storage. In large-scale laser positioning, it is only necessary to verify the feature coding within a certain range around the current position, thereby reducing the impact of reduced search efficiency caused by too large laser point cloud data, and solving the problem of too long searching for laser feature maps in large-scale laser positioning.

[0057] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A SLAM fast initial positioning method based on high-precision maps, suitable for a mobile carrier equipped with a high-precision map system and a laser positioning system, characterized in that: Obtaining the position and posture of the current position of the mobile carrier, and calculating the characteristic value of the current position; the characteristic value includes parameters that can reflect the characteristics of the landmarks within a certain range around the current position; Extract all feature codes within a certain range around the current position through the posture of the current position; the feature codes are pre-stored, contain posture information, feature values, and correspond to the laser point cloud cluster; Traversing all the feature codes, searching for a feature code that is consistent with the feature value of the current position, so as to locate the corresponding laser point cloud cluster; The corresponding laser point cloud cluster is used to complete the initialization of the laser positioning system.

2. The SLAM fast initial positioning method based on high-precision map according to claim 1, characterized in that: The feature code is obtained in the following manner: when establishing a laser point cloud feature map, the initial position and posture of the mobile carrier obtained by the combined navigation is recorded at the initial moment, and the established laser point cloud data is converted to the global coordinate system through the initial position and posture; in the process of establishing the laser point cloud feature map, the high-precision map system is continuously subscribed to through the combined navigation to obtain environmental features within a certain range around the current position; the feature values ​​in the environmental features are extracted, and the feature codes are formed with the position and posture of the current position, corresponding to the laser point cloud cluster centered on the current position.

3. The SLAM fast initial positioning method based on high-precision map according to claim 1 or 2, characterized in that: The feature code is a combination of the coordinates of the current position, the attitude angle and the feature value in a certain order.

4. The SLAM fast initial positioning method based on high-precision map according to claim 2, characterized in that: The environmental features include semantic information of lane lines and signboards.

5. The SLAM fast initial positioning method based on high-precision map according to claim 4 is characterized in that: The characteristic values ​​include the total number of lane lines, the average lane line color, the average lane line type, the total number of signboards, and the average signboard type.

6. The SLAM fast initial positioning method based on high-precision map according to claim 1, characterized in that: The characteristic value of the current position is calculated by a high-precision map system.

7. The SLAM fast initial positioning method based on high-precision map according to claim 1, characterized in that: The certain range around the current position is within a range of 8-20m around the current position.

8. The SLAM fast initial positioning method based on high-precision map according to claim 1, characterized in that: The mobile carrier is a vehicle.

9. A SLAM fast initial positioning system based on high-precision maps, characterized by: include: The current position feature value calculation module is used to obtain the position and posture of the current position from the integrated navigation carried by the mobile carrier and calculate the feature value of the current position; The characteristic value includes parameters that can reflect the characteristics of landmarks within a certain range around the current position; A feature code extraction module is used to extract all feature codes within a certain range around the current position through the posture of the current position; the feature code is pre-stored, includes posture information, feature values, and corresponds to the laser point cloud; A laser point cloud positioning module is used to traverse all the feature codes and find a feature code that is consistent with the feature value of the current position, so as to locate the corresponding laser point cloud; The initialization positioning module is used to use the corresponding laser point cloud to complete the initialization of the laser positioning system.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the SLAM fast initial positioning method based on high-precision map described in any one of claims 1-8 is implemented.

Images