Geometric feature-based real-time map matching method

Abstract

The invention discloses a real-time map matching method based on geometric features, which comprises the following steps: establishing the geometric features of a target vehicle based on the gridding processing of a high-precision map, determining the current candidate matching road set of the target vehicle according to the global UTM coordinate of the target vehicle, traversing the set, and determining the current candidate matching road set of the target vehicle; sequentially calculating the similarity between the current state of the target vehicle and all lanes in each candidate road, selecting the lane with the highest similarity as a matching result, and calculating the coordinates of the target vehicle relative to the matched road in a coordinate system established based on the road reference line according to the matching result; the distance and direction information is comprehensively considered to measure the similarity between the current state of the vehicle and the lane center line, and the map matching algorithm based on geometric features effectively reduces the range of map matching candidate lanes, so that the map matching efficiency is improved, and the real-time performance of the matching process is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving map matching and positioning, and specifically relates to a real-time map matching method based on geometric features. Background Art

[0002] Map matching algorithm is a key technology based on geographic information system (GIS) and plays a vital role in autonomous driving vehicles. Its core idea is to determine the current road, lane and position of the vehicle in the lane by comparing the vehicle's sensor data (such as GPS or lidar) with the electronic map data. Map matching algorithm has important significance in the field of autonomous driving, including the following points:

[0003] Map matching algorithms can be used to correct positioning errors of GPS or LiDAR and improve positioning accuracy. Generally speaking, vehicles usually travel on roads. However, due to factors such as bad weather or measurement errors of the positioning module itself, the positioning of the vehicle may deviate from the road. At this time, the use of map matching algorithms can effectively correct the deviation.

[0004] Map matching algorithms can significantly improve the perception capabilities of autonomous vehicles. Generally speaking, the vehicle's sensors can only obtain the vehicle's global location information (such as latitude and longitude or UTM coordinates). Map matching algorithms can help the vehicle further obtain a large amount of static scene information around it, such as lane lines, intersections, traffic lights, stop lines, etc. This information can be used to supplement sensor data and enhance the perception capabilities of autonomous vehicles.

[0005] Map matching algorithms can also provide a large amount of prior information for subsequent modules of autonomous driving, such as trajectory prediction and decision planning, to improve the intelligence of the system.

[0006] Due to the importance of map matching, the algorithm has become one of the research hotspots in the field of autonomous driving. Among them, the map matching algorithm based on geometric features has been widely studied and applied to actual production environments due to its strong interpretability, stability and low computational complexity. The core idea of ​​this method is to match the current state information of the vehicle with the center lines of all candidate lanes by similarity, and select the lane with the highest similarity as the matching result. Although the algorithm has the above advantages, when the algorithm is actually deployed on a real vehicle, since the map contains a large number of lanes in most cases, it is difficult to meet the real-time requirements of the autonomous driving system if only global map matching is performed on all lanes of the map. Summary of the invention

[0007] In view of this, the present invention provides a real-time map matching method based on geometric features, which establishes the geometric features of the target vehicle based on grid processing of the high-precision map to achieve real-time matching of the target vehicle to the map.

[0008] The present invention provides a real-time map matching method based on geometric features, which specifically includes the following steps:

[0009] Step 1: Divide the high-precision map into multiple square grids. The area composed of the grids completely covers all roads in the map. The borders of the grids are parallel to the UTM coordinate system of the map. The lower left corner of the map is used as the reference point P. r , P r The global UTM coordinates are (x r ,y r ), the grid where any point P on the map is located and its index in the grid are recorded as the grid index of P according to the reference point. The global UTM coordinates of point P are (x, y). The grid index includes the horizontal index x index and the vertical index y index They are: and d is the grid width;

[0010] Step 2: determine the grid road index table of the grid, which is used to record the numbers of all roads contained in the grid; establish a candidate matching road set of the vehicle to store optional roads matching the vehicle, and initialize the candidate matching road set to an empty set;

[0011] Step 3: Obtain the real-time UTM coordinates of the target vehicle to be matched. If the UTM coordinates exist in the candidate matching road set of the target vehicle, execute step 4; otherwise, execute step 5.

[0012] Step 4: Obtain a matching road corresponding to the UTM coordinates in the candidate matching road set, obtain the predecessor road and the successor road adjacent to the matching road from the high-precision map, and the matching road and its predecessor road and successor road constitute the current candidate matching road set corresponding to the UTM coordinates, and execute step 6;

[0013] Step 5: Calculate the grid index of the grid to which the target vehicle belongs according to the real-time UTM coordinates obtained in step 3, and save all roads contained in the grid to which the target vehicle belongs and other grids adjacent to the grid into the candidate matching road set of the UTM coordinates;

[0014] Step 6. Calculate the similarity between the center line of each lane in the candidate matching road set and the current position of the vehicle, take the lane with the highest similarity as the final matching result, obtain the lane ID corresponding to the final matching result, and then obtain the corresponding road ID based on the lane ID, convert the global UTM coordinates of the target vehicle into ST coordinates, and complete the real-time matching of the vehicle and the map. The ST coordinate system is based on the road reference line, the direction along the road reference line is the S direction, and the direction perpendicular to the road reference line is the T direction.

[0015] Furthermore, the method of determining the grid road index table of the grid is: using the horizontal index and vertical index calculation formula to sequentially calculate the grid index to which each point on all road reference lines in the map belongs, if a discrete point on a road reference line is located in a certain grid, then it is determined that the grid contains the road, and the grid road index table of the grid is obtained, in which the road IDs of all roads contained in each grid unit are recorded.

[0016] Furthermore, the grid road index table is obtained in an offline manner and saved in the form of a file.

[0017] Furthermore, in step 6, the similarity between the center line of each lane in the candidate matching road set and the current position of the vehicle is calculated, and the similarity function used is shown in the following formula:

[0018] S i =w1Δd i +w2Δθ i

[0019] Among them, S i is the similarity between the current position of the target vehicle and the centerline of the i-th lane in the candidate matching road set, Δd i is the vertical distance between the current position of the target vehicle and the projection point of the lane centerline, Δθ i is the heading deviation between the current position of the target vehicle and the projection point of the lane centerline. w1 and w2 are both weight coefficients. The projection point is the point closest to the current position of the target vehicle among all discrete points on the lane centerline.

[0020] Furthermore, the projection point is obtained by binary search.

[0021] Furthermore, the method of obtaining the corresponding road ID according to the lane ID in step 6 is: querying the road ID to which the lane ID belongs in the OpenDrive map according to the lane ID.

[0022] Beneficial effects:

[0023] The present invention establishes geometric features of a target vehicle based on gridding processing of a high-precision map, determines the current candidate matching road set of the target vehicle according to the global UTM coordinates of the target vehicle, traverses the set, calculates the similarity between the current state of the target vehicle and all lanes in each candidate road in turn, selects the lane with the highest similarity as the matching result, and then calculates the coordinates of the target vehicle relative to the matching road in a coordinate system established based on a road reference line according to the matching result, comprehensively considers two types of information, distance and direction, to measure the similarity between the current state of the vehicle and the center line of the lane, and the map matching algorithm based on geometric features effectively narrows the range of candidate lanes for map matching, thereby improving the efficiency of map matching and ensuring the real-time performance of the matching process. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the processing flow of a real-time map matching method based on geometric features provided by the present invention.

[0025] Figure 2 A schematic diagram of a high-precision map after gridding using a real-time map matching method based on geometric features provided by the present invention.

[0026] Figure 3 A schematic diagram of variables involved in a similarity function used in a real-time map matching method based on geometric features provided by the present invention.

[0027] Figure 4 A schematic diagram of converting UTM coordinates to ST coordinates used in a real-time map matching method based on geometric features provided by the present invention. DETAILED DESCRIPTION

[0028] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0029] The present invention provides a real-time map matching method based on geometric features, the basic idea of ​​which is: determine the current candidate matching road set of the target vehicle according to the global UTM coordinates of the target vehicle, traverse the set, calculate the similarity between the current state of the target vehicle and all lanes in each candidate road in turn, select the lane with the highest similarity as the matching result, and then calculate the coordinates of the target vehicle relative to the matching road in a coordinate system established based on a road reference line according to the matching result.

[0030] The map used in the present invention is a high-precision map in the OpenDrive format, and the map in this format contains the centerline information of all lanes. The ST coordinate system is the coordinate system relative to the road reference line in the OpenDrive map, where the S direction is along the road reference line and the T direction is perpendicular to the road reference line. Existing map matching algorithms usually rely on global search, or perform complex calculations through a large amount of sensor data to determine the matching relationship between the vehicle and the road, resulting in low matching efficiency, especially in autonomous driving scenarios with high real-time requirements. Therefore, in order to improve the real-time performance of map matching, the present invention proposes a real-time map matching strategy based on a map gridding strategy to determine a set of candidate roads, so as to narrow the scope of matching roads and improve the efficiency and accuracy of the algorithm.

[0031] The present invention provides a real-time map matching method based on geometric features, the overall process is as follows: Figure 1 As shown, the specific steps include:

[0032] Step 1: Grid the high-precision map, that is, divide the high-precision map into multiple grids, such as Figure 2 As shown, the grids are all squares. The area composed of all grids can completely cover all roads in the map. The borders of the grids are parallel to the UTM coordinate system of the map. The lower left corner of the map is used as the reference point P r , P r The global UTM coordinates are (x r ,y r ), for any point P on the map, the grid where the point is located and its index in the grid can be determined according to the reference point, which is recorded as the grid index of P. The grid index includes the horizontal index and the vertical index, which are respectively:

[0033] Horizontal Index

[0034] Vertical Index

[0035] Among them, (x, y) is the global UTM coordinate of point P, d is the grid width, and the index of the grid to which each point on all road reference lines belongs is obtained; if there is a discrete point located in a certain grid on the road reference line, it is considered that the grid contains the road, and the grid road index table of each grid is obtained, and the grid road index table records the numbers of all roads contained in each grid. A candidate matching road set for the vehicle is established, which saves the candidate matching roads of the optional roads according to the current position of the vehicle, and the candidate matching road set is initialized to an empty set.

[0036] The method for obtaining the grid index of the grid to which each point on all road reference lines belongs is: each point on all road reference lines in the map is sequentially substituted into the horizontal index and vertical index calculation formula to calculate the grid index to which it belongs. If a discrete point on a road reference line is located in a grid, it is considered that the grid contains the road. The grid road index table is obtained through the above operation, which records the road IDs of all roads contained in each grid unit.

[0037] To further improve the real-time performance of the algorithm, the grid road index table is obtained offline and saved in the form of a file to reduce its impact on the operation of each cycle of the autonomous driving system.

[0038] Step 2: Obtain the real-time UTM coordinates of the target vehicle to be matched. If the UTM coordinates exist in the candidate matching road set, execute step 3; otherwise, execute step 4.

[0039] Step 3: Obtain the matching road corresponding to the UTM coordinates in the candidate matching road set, obtain the predecessor road and successor road adjacent to the matching road from the high-precision map based on the matching road, and the matching road and its predecessor road and successor road constitute the current candidate matching road set corresponding to the UTM coordinates, and execute step 5.

[0040] The present invention constructs a strategy for candidate matching roads based on historical information, which can effectively narrow the matching range and improve the efficiency of the algorithm. Specifically, when the target vehicle has been detected by the sensor before, the algorithm can use the result of the previous map matching to optimize the selection of candidate roads. Since the detection frequency of the sensor is high and the movement range of the vehicle is limited during the continuous detection cycle, the algorithm can use the matching result of the previous frame as a reference and use the road matched by the vehicle in the previous frame and its successor roads as the current candidate road set.

[0041] Step 4: Calculate the grid index of the grid to which the target vehicle belongs according to the real-time UTM coordinates of the target vehicle to be matched obtained in step 2, and save all roads contained in the grid to which it belongs and the eight grids adjacent to the grid into the candidate matching road set of the UTM coordinates, such as Figure 2 The dotted area is shown.

[0042] Since the grid road index table obtained in advance by the algorithm is not completely accurate, some grid units may be omitted. Here, the present invention expands the roads contained in the surrounding adjacent grids to the candidate matching road set based on the grid to which the target vehicle belongs to contain roads. In addition, since the actual road is a two-dimensional entity with width, and the road reference line is a one-dimensional entity with only length, it is inaccurate to indirectly judge which grids the road belongs to based on the discrete road reference line coordinate points. Therefore, the present invention also needs to appropriately expand the matching grid area to compensate for this problem.

[0043] Step 5: Calculate the similarity between the center line of each lane in the candidate matching road set and the current position of the vehicle. The similarity matching involves variables such as Figure 3 As shown, the lane with the highest similarity is taken as the final matching result, the lane ID corresponding to the final matching result is obtained, and then the corresponding road ID is obtained according to the lane ID.

[0044] In order to ensure the accuracy of the measurement, the present invention adopts a similarity function for measurement, and the similarity function is shown in the following formula:

[0045] S i =w1Δd i +w2Δθ i (3)

[0046] Among them, S i is the similarity between the current position of the target vehicle and the centerline of the i-th lane in the candidate matching road set, Δd i is the vertical distance between the current position of the target vehicle and the projection point of the lane centerline, Δθ i is the heading deviation between the current position of the target vehicle and the projection point on the center line of the lane, w1 and w2 are weight coefficients, and the projection point is the point closest to the current position of the target vehicle among all discrete points on the center line of the lane. According to the definition of formula (3), S i The smaller the value, the higher the similarity between the vehicle's current position and the lane.

[0047] In order to ensure the accuracy of the map matching algorithm in a more complex lane overlap environment (intersection scenario), formula (3) comprehensively considers the distance deviation factor and the heading deviation factor when measuring the similarity between the vehicle and the lane. In addition, the projection point can be obtained through binary search, such as Figure 4 shown.

[0048] The process of selecting the lane with the highest similarity as the matching result can be expressed as follows:

[0049]

[0050] Among them, i ★ is the final matching result, which means the lane ID with the highest similarity to the vehicle among all lanes in the candidate matching road set.

[0051] Furthermore, the lane ID can be further queried in the OpenDrive map to obtain the road ID to which it belongs.

[0052] Step 6: Convert the global UTM coordinates of the target vehicle into ST coordinates.

[0053] In order to facilitate the calculation of subsequent modules of the autonomous driving system (such as prediction, planning, decision-making, etc.), after obtaining the lane ID and road ID to which the vehicle belongs, the map matching algorithm needs to further convert the vehicle's global UTM coordinates into the ST coordinate system. The ST coordinate system is essentially a Frenet coordinate system. In the OpenDrive format high-precision map, the ST coordinate system is based on the road reference line, with the direction along the road reference line as the S direction and the direction perpendicular to the road reference line as the T direction. The conversion process is shown in formulas (5) and (6):

[0054]

[0055] Among them, p is the target vehicle position to be converted, and its UTM coordinates (x, y) are known. Point p1 is the projection point of p on the center line of its lane. The UTM coordinates (x1, y1) and ST coordinates (s1, t1) of this point can be parsed from the OpenDrive high-precision map as known quantities. is the vector from the projection point to the vehicle position. Since the UTM coordinates of the projection point p1 and the vehicle position p are known, this vector is also known and can be expressed as (x-x1, y-y1); is the unit direction vector at the projection point, which can be parsed from the OpenDrive map.

[0056] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A real-time map matching method based on geometric features, characterized in that: The specific steps include: Step 1: Divide the high-precision map into multiple square grids. The area composed of the grids completely covers all roads in the map. The borders of the grids are parallel to the UTM coordinate system of the map. The lower left corner of the map is used as the reference point P. r , P r The global UTM coordinates of point P on the map are (xr, yr). The grid where any point P on the map is located and its index in the grid are recorded as the grid index of P. The global UTM coordinates of point P are (x, y). The grid index includes the horizontal index x index and the vertical index y index They are: and d is the grid width; Step 2: determine the grid road index table of the grid, which is used to record the numbers of all roads contained in the grid; establish a candidate matching road set of the vehicle to store optional roads matching the vehicle, and initialize the candidate matching road set to an empty set; Step 3: Obtain the real-time UTM coordinates of the target vehicle to be matched. If the UTM coordinates exist in the candidate matching road set of the target vehicle, execute step 4; otherwise, execute step 5. Step 4: Obtain a matching road corresponding to the UTM coordinates in the candidate matching road set, obtain the predecessor road and the successor road adjacent to the matching road from the high-precision map, and the matching road and its predecessor road and successor road constitute the current candidate matching road set corresponding to the UTM coordinates, and execute step 6; Step 5: Calculate the grid index of the grid to which the target vehicle belongs according to the real-time UTM coordinates obtained in step 3, and save all roads contained in the grid to which the target vehicle belongs and other grids adjacent to the grid into the candidate matching road set of the UTM coordinates; Step 6. Calculate the similarity between the center line of each lane in the candidate matching road set and the current position of the vehicle, take the lane with the highest similarity as the final matching result, obtain the lane ID corresponding to the final matching result, and then obtain the corresponding road ID based on the lane ID, convert the global UTM coordinates of the target vehicle into ST coordinates, and complete the real-time matching of the vehicle and the map. The ST coordinate system is based on the road reference line, the direction along the road reference line is the S direction, and the direction perpendicular to the road reference line is the T direction.

2. The real-time map matching method according to claim 1, characterized in that: The method for determining the grid road index table of the grid is: using the horizontal index and vertical index calculation formula to sequentially calculate the grid index to which each point on all road reference lines in the map belongs, if a discrete point on a road reference line is located in a grid, it is determined that the grid contains the road, and the grid road index table of the grid is obtained, in which the road IDs of all roads contained in each grid unit are recorded.

3. The real-time map matching method according to claim 1, characterized in that: The grid road index table is obtained in an offline manner and saved in the form of a file.

4. The real-time map matching method according to claim 1, characterized in that: In step 6, the similarity between the center line of each lane in the candidate matching road set and the current position of the vehicle is calculated, and the similarity function used is shown in the following formula: S i =w1Δd i +w2Δθ i Among them, S i is the similarity between the current position of the target vehicle and the centerline of the i-th lane in the candidate matching road set, Δd i is the vertical distance between the current position of the target vehicle and the projection point of the lane centerline, Δθ i is the heading deviation between the current position of the target vehicle and the projection point of the lane centerline. w1 and w2 are both weight coefficients. The projection point is the point closest to the current position of the target vehicle among all discrete points on the lane centerline.

5. The real-time map matching method according to claim 4, characterized in that: The projection point is obtained by binary search.

6. The real-time map matching method according to claim 1, characterized in that: The method of obtaining the corresponding road ID according to the lane ID in step 6 is: querying the road ID to which the lane ID belongs in the OpenDrive map according to the lane ID.

Images