Road topology construction method and system and medium

Abstract

The invention relates to a road topology construction method and system and a medium, and relates to the field of map construction, and the method comprises the steps: mapping a sensing lane line to a sampling grid chart; carrying out iterative search on an area covered by the lane lines to identify topological nodes; recording topology change conditions and topology boundary information after each search; and after the end of the lane line is searched, sequentially connecting the topological nodes according to the recorded topological change sequence to construct a complete road topology. Complete road topology information can be constructed online only through the lane lines output by the sensing module without depending on a high-precision map.

Description

Technical Field

[0001] The present invention relates to the field of map construction, and in particular to a road topology construction method, system and medium. Background Art

[0002] Road topology information plays an important role in the decision-making and planning modules of the autonomous driving system, and is usually provided by the high-precision map module. However, the road topology output by the high-precision map has problems such as slow update cycle and inaccurate information. Especially in scenarios such as temporary road repairs, the topology information provided by the high-precision map is often inconsistent with the actual situation. There are currently two mainstream processing methods in the industry: one is to use the lane lines identified by the perception module to verify the road topology of the high-precision map; the other is to construct the topology of the lane where the vehicle is located only by identifying the perception lane lines on the left and right of the vehicle.

[0003] These existing methods all have certain limitations. Although the first method can identify situations where the HD map topology does not match the actual situation to a certain extent, it cannot correct the wrong road topology and can only choose to downgrade or exit the function, affecting the driving experience. Although the second method does not rely on HD maps, it can only construct the topology of the lane where the vehicle is located, and cannot provide topological information of other lanes. It is difficult to meet the needs of the decision-making and planning modules of the autonomous driving system in complex scenarios.

[0004] Therefore, there is an urgent need for a method that can accurately construct the complete road topology in real time to improve the adaptability and reliability of the autonomous driving system in various complex scenarios. Summary of the invention

[0005] In view of the shortcomings of the prior art mentioned above, the present invention provides a road topology construction method, system and medium, which aim to construct complete road topology information online without relying on high-precision maps, only through the lane lines output by the perception module.

[0006] To achieve the above purpose, the present invention adopts the following technical solution.

[0007] In a first aspect, the present invention provides a road topology construction method, which adopts the following technical solution: Mapping the perceived lane lines to the sampling grid map; Iteratively searching the area covered by the lane line to identify topological nodes; Record the topology changes and topology boundary information after each search; After searching to the end of the lane line, the topological nodes are connected in sequence according to the recorded topological change order to construct a complete road topology.

[0008] The iterative search of the area covered by the lane line includes: Initialize the particles of the particle filter; Perform a single-step search to update the particle position; Make a line crossing judgment and delete the line crossing particles; The remaining particles are clustered to generate topological nodes.

[0009] In a possible implementation, the particles of the initialized particle filter include: Dynamically adjust the number of particles according to the number of lane lines; The initial position coordinates of the particles are determined by random sampling near the starting position of the lane line.

[0010] In a possible implementation, the single-step search and updating of the position of the particle includes: Updates the particle's position coordinates along its current direction according to the preset step size.

[0011] In a possible implementation, the performing line crossing determination includes: Map the position coordinates of the particles before and after the update to the occupancy grid map; Check whether there is a grid occupied by lane lines between the two grids after mapping.

[0012] In a possible implementation, recording the topology change includes: Determine the change in the number of topological nodes; Record the parent-child relationship of topological nodes.

[0013] In a possible implementation, recording topology boundary information includes: Starting from the topological node position, traverse the occupied grid graph perpendicular to the search direction; Record the lane line mark corresponding to the first grid occupied by the lane line encountered during the traversal process.

[0014] In a possible implementation, the method further includes: Importance resampling of the remaining particles is performed, including: Calculate the distance from each particle to the nearest lane line; Calculate the importance weight of particles based on distance; Resample according to importance weights.

[0015] In a possible implementation, resampling according to importance weights includes: Keep the position coordinates of the resampled particles unchanged; Makes a random correction to the search direction of the resampled particles.

[0016] In a second aspect, the present invention provides a road topology construction system, which adopts the following technical solution: A lane mapping module, at least for mapping the sensed lane lines to a sampling grid map; A topology search module, at least for iteratively searching the area covered by the lane line to identify topological nodes, and recording topological changes and topological boundary information after each search; The topology splicing module is used to connect the topology nodes in sequence according to the recorded topology change order to construct a complete road topology after searching to the end of the lane line.

[0017] In a third aspect, the present invention provides a readable storage medium, which adopts the following technical solution: A readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, a road topology construction method as described in any one of the first aspects above is implemented.

[0018] In summary, compared with the prior art, the present invention includes at least one of the following beneficial technical effects: The present invention solves the problem of relying on high-precision maps or only being able to build single-lane topology in the prior art by mapping the perceived lane lines to a sampling grid map and iteratively searching the coverage area to identify topological nodes. The present invention records topological changes and boundary information, builds a complete multi-lane topology online, and provides comprehensive, real-time, and accurate road topology information for the decision-making and planning of the autonomous driving system in complex scenarios, significantly improving the adaptability and safety of the autonomous driving system, avoiding functional degradation or exit due to untimely updates of high-precision maps, and optimizing the driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 It is a flowchart of a specific embodiment of a road topology construction method of the present invention.

[0021] Figure 2 It is a flow chart of a specific embodiment of a road topology construction method of the present invention.

[0022] Figure 3 is a flowchart of a specific embodiment of the topology search step of the present invention

[0023] Figure 4 It is a schematic diagram of judging the topological change of the connected area of ​​the present invention.

[0024] Figure 5 This is a schematic diagram of the construction process and effect of the road topology of the present invention.

[0025] Figure 6 It is a structural schematic diagram of a specific embodiment of a road topology construction system of the present invention. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application. In addition, it should be understood that the specific implementation methods described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0027] It should be noted that the description order of the following embodiments is not intended to limit the preferred order of the embodiments of the present application. In addition, in the following embodiments, the description of each embodiment has its own emphasis, and for parts not described in detail in one embodiment, reference can be made to the relevant description of other embodiments.

[0028] The method steps described in the embodiments of the present invention may be executed in the order described in the specific implementation manner, or the execution order of each step may be adjusted according to actual needs and on the premise that the technical problem can be solved, which will not be listed here one by one.

[0029] Reference Figure 1 and Figure 2 , an embodiment of the present invention provides a road topology construction method, comprising the following steps.

[0030] S1, maps the perceived lane lines to the sampling grid map.

[0031] Specifically, the lane lines output by the perception module are mapped to a sampling grid map, which can be a raster map, an occupied grid map, or other sampling graphs that can support topological search.

[0032] Taking the occupancy grid map as an example, its construction method includes: defining a rectangular area covering a certain distance in front and behind the vehicle position and a certain width on the left and right (for example, 200 meters × 40 meters), dividing it into grid units (for example, each unit is 0.5 meters × 0.5 meters), and recording the occupancy status of the lane line and the corresponding lane line number in the grid map.

[0033] In addition, the type and resolution of the sampling grid map can be dynamically adjusted according to the actual application requirements, such as using a higher resolution grid map to improve topological accuracy, or using a low resolution grid map suitable for computing resources to reduce computing overhead.

[0034] Lane mapping is achieved through these methods, providing a data basis for subsequent topology search.

[0035] S2, iteratively searching the area covered by the lane line to identify topological nodes.

[0036] Specifically, an iterative search is performed on the area covered by the lane lines in the sampling grid map to identify topological nodes. The search method may adopt particle filtering, A algorithm, hybrid A algorithm or heuristic search.

[0037] Taking particle filtering as an example, the number of particles is dynamically adjusted according to the number of lane lines during initialization, initial particles are randomly sampled from the vicinity of the starting position of the lane line, and the search is performed by iteratively updating the particle state (including position coordinates and search direction).

[0038] During the search, a fixed step size can be set to gradually update the particle position, and in each search, check whether the particle crosses the line (i.e., whether it crosses the lane line boundary). If so, the particle is deleted. If it does not cross the line, the particles in the same connected area are clustered to form a topological node.

[0039] In addition, to improve efficiency, in some embodiments, when the number of particles is insufficient, resampling can be used to supplement it, thereby supporting the subsequent search process.

[0040] Different search methods can be selected according to specific scenarios to ensure accurate identification and real-time processing capabilities of topological nodes.

[0041] S3, records the topology changes and topology boundary information after each search.

[0042] Specifically, in each search process, the topological changes and boundary information are recorded by analyzing the clustering results of particles and the changes in connected areas.

[0043] Topological changes can be identified by judging the increase or decrease of particle clusters and changes in connectivity, including scenarios such as unchanged topology, topological bifurcation, topological merger, and topological extinction.

[0044] While identifying topological changes, the status of each topological node (such as position coordinates) and the lane line information on the left and right boundaries are recorded.

[0045] The boundary information can be recorded by traversing the grids within a certain range in the left and right directions at the node position, marking the grids intersecting the lane lines and their corresponding lane line numbers.

[0046] In addition, to ensure the continuity and accuracy of the records, each search result can be compared with the previous round of search results to update and improve the description of topological changes, ultimately providing complete and accurate topological sequence and boundary data for subsequent topological splicing.

[0047] S4, after searching to the end of the lane line, connect the topological nodes in sequence according to the recorded topological change order to build a complete road topology.

[0048] Specifically, after completing the iterative search for lane lines, the topological nodes are connected in sequence from parent nodes to child nodes according to the recorded topological changes to construct a complete road topology.

[0049] The specific method of topological splicing is to traverse each topological node and its associated child nodes to gradually establish the connectivity information between nodes.

[0050] During the stitching process, the left and right boundary information of each node is simultaneously associated with the corresponding lane line to ensure the integrity of the topology and the accuracy of the boundary.

[0051] The stitching result can be expressed in the form of a graph structure, where each node contains its location information, connection relationship and boundary attributes.

[0052] In addition, to adapt to the needs of different scenarios, the stitching algorithm can be dynamically optimized. For example, it can adjust the connection strategy when the road is bifurcated, merged, or lane lines are missing, ensuring that the final generated road topology can truly reflect the road structure and provide comprehensive support for the decision-making and planning module of the autonomous driving system.

[0053] The road topology construction method described in the embodiment of the present invention maps the perceived lane lines to a sampling grid map, iteratively searches the coverage area to identify topological nodes, records topological changes and boundary information, and splices the topological nodes in the recorded order to construct a complete road topology, thereby achieving the ability to construct a multi-lane topology online, in real time, and accurately.

[0054] This method gets rid of the dependence on high-precision maps and effectively solves the problem of slow and incomplete topology information update in existing technologies. Even in complex scenarios such as road bifurcations, lane merging, or partial loss of perceived lane lines, it can still provide complete and reliable road topology information, providing comprehensive support for the decision-making and planning modules of the autonomous driving system, significantly improving the adaptability and safety of autonomous driving.

[0055] Further, as an embodiment of the present invention, refer to Figure 3 In step S2, iterative search of the area covered by the lane line includes: initializing particles of the particle filter; performing single-step search to update the position of the particles; making a line crossing judgment and deleting the line crossing particles; and clustering the remaining particles to generate topological nodes.

[0056] Specifically, in step S2, the iterative search of the lane line coverage area first initializes the particles of the particle filter, and the number of particles is dynamically adjusted according to the number of lane lines. The initial state is generated by random sampling near the starting position of the lane line; then the particles are searched in a single step, and the position coordinates and direction of the particles are updated by setting the step size; after each update, the crossing line judgment is performed by checking whether the position of the particle before and after the update crosses the lane line boundary. If so, the particle is deleted; for the particles that are not deleted, the clustering algorithm is used to classify the particles in the same connected area into one category, thereby generating the corresponding topological nodes. This process can be achieved by mapping to the sampling grid graph to ensure the accuracy and connectivity of the topological search, and provide basic support for the subsequent topological change identification and splicing.

[0057] Furthermore, as an implementation mode of the present invention, the particles of the initialization particle filter include: dynamically adjusting the number of particles according to the number of lane lines; and randomly sampling near the starting position of the lane line to determine the initial position coordinates of the particles.

[0058] Specifically, topology search is an iterative search process. In the first search, the particles of the particle filter need to be initialized. The number of particles n can be dynamically adjusted according to the number of lane lines: n=k1*nLine Among them, k1 is the adjustment factor and nLine is the number of lane lines.

[0059] The state of each particle can be expressed as s i ={dx i ,dy i ,alp i}, where dx i ,dy i ,alp i Represent the longitudinal position coordinates, lateral position coordinates and search direction angle of the particle respectively. Start searching the topology from the starting position of each lane line, then dx i ,dy i It can be determined by randomly sampling coordinate points near the starting position of the lane line. i It can be obtained by random sampling from a standard normal distribution.

[0060] If the current search phase is not the first search, no initialization is required, and the state of each particle is the particle state after the last resampling of the previous search.

[0061] Further, as an implementation mode of the present invention, the single-step search and updating of the position of the particle includes: updating the position coordinates of the particle along the current direction of the particle according to a preset step length.

[0062] Specifically, the single-step search process is achieved by updating the position coordinates of all particles. The update step size can be an adjustable parameter, set to dStep. Figure 4 (a) shows a typical single-step search process in a scenario where the topology does not change. For the i-th particle, the state after the j-th search is Then in the j+1th search, the particle moves along The direction updates the coordinate position according to the step size dStep:

[0063] Furthermore, as an embodiment of the present invention, the lane crossing judgment includes: mapping the position coordinates of the particles before and after the update to an occupied grid map; and checking whether there is a grid occupied by a lane line between the two grids after the mapping.

[0064] Specifically, check whether each particle crosses any lane line after being updated. If so, delete it. Figure 4 (d) Update the two position coordinates before and after Map it to the occupied grid map, and then traverse all the grids between the two mapped grids. If there is a grid occupied by the lane line, it means that the particle has crossed the line before and after the update, and the particle is deleted.

[0065] By traversing all particles and clustering the particles in the same connected region into one category, the topological nodes of the connected region can be obtained. The specific method is to map the position coordinates of the particles that have been processed after crossing the lane to the occupied grid map. If there is no grid occupied by lane lines between particles, these particles can be clustered into the same category. Figure 4 (b) and Figure 4 (c) shows the clustering results in different topological scenarios.

[0066] Further, as an implementation of the present invention, in step S3, recording the topology change includes: determining the change in the number of topology nodes; and recording the parent-child relationship of the topology nodes.

[0067] Specifically, the clustering results described in the above embodiment can determine the topological changes of each single-step search, record the changes and output them to the subsequent modules. Assume that after the j-th search clustering, a total of M topological nodes are obtained, and the m-th topological node is recorded as Then in the j+1th search process, the components After all particles of are processed in step S2 in sequence, there may be Figure 4 Four situations are shown.

[0068] like Figure 4 As shown in (a), if the composition All particles in have only one topological node after the j+1th search, which is recorded as And composition The particles come from but and If they belong to the same topology, the topology does not change from the jth search to the j+1th search. The topology in this scenario can be written as: in, Represents a topological node The child nodes of Represents a topological node The parent node of .

[0069] like Figure 4 As shown in (b), if the composition All particles have only one topological node after the j+1th search and The particles come from and in, Indicates that after the jth search, For other topological nodes except , there is a topological merge in the process from the jth search to the j+1th search. The topological change in this scenario can be recorded as:

[0070] like Figure 4 As shown in (c), if the composition After the j+1th search, all particles in There is a topological bifurcation in the process from the jth search to the j+1th search. The topological change in this scenario can be recorded as:

[0071] like Figure 4 As shown in (d), if the composition All particles of do not have any topological nodes after the j+1th search, that is If all particles in are deleted after the jth search, the topology will disappear in the j+1th search. The topological changes in this scenario can be recorded as:

[0072] Furthermore, as an embodiment of the present invention, in step S3, recording the topological boundary information includes: starting from the topological node position, traversing the occupied grid map along a direction perpendicular to the search direction; recording the lane line logo corresponding to the first grid occupied by the lane line encountered during the traversal process.

[0073] Specifically, topological nodes The state of is the average of the states of all particles that make up the cluster:

[0074] Among them, Location coordinates Mapped to the occupied grid map, and taking this coordinate as the starting point, respectively along The grids within a certain distance are traversed in the direction. If there is a grid occupied by a lane line, the lane line id corresponding to the grid is recorded.

[0075] like If there is a lane line in the direction, it is considered a topological node The left boundary of the lane line is recorded in middle.

[0076] like If there is a lane line with id line1 in the direction, it will be regarded as a topological node The right boundary of the lane line is recorded in middle.

[0077] If no grid occupied by lane lines is identified after traversing a certain distance, it means No left or right border.

[0078] Furthermore, as an embodiment of the present invention, the road topology construction method also includes importance resampling of the remaining particles.

[0079] The importance resampling of the remaining particles includes: calculating the distance from each particle to the nearest lane line; calculating the importance weight of the particle according to the distance; and resampling according to the importance weight.

[0080] The resampling according to the importance weight includes: keeping the position coordinates of the resampled particles unchanged; and randomly correcting the search direction of the resampled particles.

[0081] Specifically, since there is an operation of deleting particles in the line crossing judgment, if resampling is not performed, the number of particles will gradually decrease and cannot support subsequent searches. The present invention solves this problem by resampling. First, the importance of each particle is calculated based on the distance from the nearest lane line. The farther the particle is from the nearest lane line, the greater its importance. Then resampling is performed based on the importance. The more important the particle is, the greater the probability of being resampled.

[0082] First, calculate the importance weight. represents the distance from the ith particle to the kth lane line after the jth search, then the importance weight of the ith particle after the jth search is Inversely proportional to its distance to the nearest lane line:

[0083] Next, normalize the importance weights of all particles in this search:

[0084] At the end of the jth search, all particles are randomly sampled and the normalized weight of each particle is That is the probability of being sampled. After resampling to obtain n particles, keep unchanged, A random correction is made based on the original value. The correction method can be: Sample from a Gaussian distribution with the mean: Thus, we get the state of the i-th particle after the j-th search:

[0085] Finally, it is determined whether all particles are still within the occupied grid range. If so, iterative steps S2 and S3 are repeated until the search is outside the occupied grid range.

[0086] Further, as an implementation of the present invention, step S4, after searching to the end of the lane line, sequentially connecting the topological nodes according to the recorded topological change order to construct a complete road topology, specifically includes: Starting from the starting topology node, connecting each node in the search module in the order of parent node to child node, a complete topology can be obtained. The left and right boundaries of the topology are the lane lines corresponding to the left and right boundaries of each node. All the constructed lane topologies together constitute the complete road topology of the current scene, which is output as a result to other modules of the autonomous driving system.

[0087] Through the analysis of a large amount of road test data, the present invention can better construct accurate and reliable road topology. Figure 5Demonstrates the process of constructing road topology by sensing lane lines.

[0088] Figure 5 (a) shows the original lane line perceived in a complex scenario, whose topology includes various complex scenarios such as road bifurcation, disappearance, and loss of perceived lane lines.

[0089] Figure 5 (b) shows the result of mapping the perceived lane lines to the occupied grid map. For easy visualization, Figure 5 In (b), the virtual and real line types of the lane lines are represented by grids of different grayscales. It is not necessary to distinguish between the virtual and real line types during the actual operation of the algorithm.

[0090] Figure 5 (c) shows the topology search process for a road bifurcation scenario. The search is performed from left to right, and the particles used for the search in the first section of the topology are represented by "○". When the search reaches the road bifurcation point, the location is accurately determined to be a bifurcation scene according to the method described in the topology change identification of the present invention, and the particles used for subsequent searches in the bifurcation topology are represented by "╳" and "▽" respectively.

[0091] Figure 5 (d) shows the construction effect of the fork topology. It performs parent node->child node splicing of the topological nodes of the fork scene according to the method in the topology splicing module of the present invention, and connects the left and right boundaries, and finally obtains the forked road topology.

[0092] like Figure 5 As shown in (e), the present invention can accurately construct the disappearing lane topology.

[0093] Figure 5 (f) The topology search results shown indicate that even if part of the perceived lane lines are lost, the present invention can still accurately describe the road topology.

[0094] It can be seen that the present invention can better construct accurate and reliable road topology by sensing lane lines, and effectively help the autonomous driving system get rid of its dependence on high-precision map topology.

[0095] The embodiment of the present invention also discloses a road topology construction system.

[0096] Reference Figure 6 A road topology construction system includes a lane mapping module 1, a topology search module 2 and a topology splicing module 3.

[0097] The lane mapping module 1 receives the lane line information output by the perception module and maps it to a sampling grid map. The sampling grid map can be a raster map, an occupied grid map, or other two-dimensional graph structures suitable for topological search. Taking the occupied grid map as an example, first define a rectangular area covering a certain range of the vehicle position (for example, 200 meters front and back, 40 meters left and right), and then divide the area into fixed-size grid units (such as 0.5 meters × 0.5 meters) to construct a two-dimensional grid map.

[0098] During the mapping process, each grid cell covered by a lane line is marked as "occupied" and its corresponding lane line number is recorded to distinguish different lane lines in subsequent searches.

[0099] The lane mapping module 1 can also dynamically adjust the resolution of the grid map to strike a balance between accuracy and computational efficiency.

[0100] In this way, the lane mapping module 1 provides a visualization and sampling basic data structure for subsequent topology search and construction.

[0101] In the sampling grid map, the topology search module 2 iteratively searches the area covered by the perceived lane lines to identify topological nodes. First, by initializing the particle filter, the number of particles is dynamically adjusted according to the number of lane lines, and the initial position and direction of the particles are randomly generated; then a single-step search is performed to update the position coordinates and direction of the particles according to the set step size. During the search process, the particles are judged to cross the lane line. If the particle position crosses the lane line boundary after the update, it will be deleted; for the particles that are not deleted, the clustering algorithm is used to cluster the particles in the same connected area into a topological node.

[0102] After each search, the topology search module 2 records the changes of the topology nodes, including information such as the topology unchanged, bifurcated, merged or disappeared, and identifies the left and right boundary lanes of the nodes. By repeating the search until the search of the coverage area is completed, the topology splicing module 3 is provided with complete topology node data and change information.

[0103] The topology splicing module 3 receives the topology nodes and their change information output by the topology search module 2, and connects all topology nodes in the order of the recorded parent node to the child node to construct a complete road topology. During the splicing process, the topology splicing module 3 associates the corresponding lane line to the topology boundary according to the left and right boundary information of each topology node, ensuring that the generated road topology not only has connectivity, but also accurately reflects the boundary structure of the road.

[0104] For scenarios involving complex topological changes such as forks, merges, or extinctions, the topological splicing module 3 dynamically adjusts the splicing logic to ensure that the connection sequence is correct and the topological structure is complete.

[0105] After stitching is completed, the topological structure of all lanes is integrated into a complete road topology map of the current scene and output to the decision-making planning module of the autonomous driving system to support real-time driving decisions and path planning.

[0106] The embodiment of the present invention also discloses a readable storage medium.

[0107] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the road topology construction method described in any one of the above embodiments. The computer-readable storage medium may include: any entity or device capable of carrying a computer program, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc. The computer program includes a computer program code. The computer program code may be in source code form, object code form, an executable file, or some intermediate form, etc. The computer-readable storage medium may include: any entity or device capable of carrying a computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc.

[0108] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0109] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processing module, or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0110] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A road topology construction method, characterized in that: include: Map the perceived lane lines to the sampling grid map; Iteratively searching the area covered by the lane line to identify topological nodes; Record the topology changes and topology boundary information after each search; After searching to the end of the lane line, the topological nodes are connected in sequence according to the recorded topological change order to construct a complete road topology.

2. The road topology construction method according to claim 1, characterized in that: The iterative search of the area covered by the lane line includes: Initialize the particles of the particle filter; Perform a single-step search to update the particle position; Make a line crossing judgment and delete the line crossing particles; The remaining particles are clustered to generate topological nodes.

3. The road topology construction method according to claim 2, characterized in that: The particles of the initialized particle filter include: Dynamically adjust the number of particles according to the number of lane lines; The initial position coordinates of the particles are determined by random sampling near the starting position of the lane line.

4. The road topology construction method according to claim 2, characterized in that: The single-step search and update position of particles comprises: Updates the particle's position coordinates along its current direction according to the preset step size.

5. The road topology construction method according to claim 2, characterized in that: The line crossing judgment comprises: Map the position coordinates of the particles before and after the update to the occupancy grid map; Check whether there is a grid occupied by lane lines between the two grids after mapping.

6. The road topology construction method according to claim 1, characterized in that: The recording of topology changes includes: determining the change in the number of topology nodes; Record the parent-child relationship of topological nodes.

7. The road topology construction method according to claim 1, characterized in that: The recording of topological boundary information includes: starting from the topological node position, traversing the occupied grid graph along a direction perpendicular to the search direction; Record the lane line mark corresponding to the first grid occupied by the lane line encountered during the traversal process.

8. The road topology construction method according to claim 1, characterized in that: Also includes: Importance resampling of the remaining particles is performed, including: Calculate the distance from each particle to the nearest lane line; Calculate the importance weight of particles based on distance; Resample according to importance weights.

9. The road topology construction method according to claim 8, characterized in that: The resampling according to importance weights includes: Keep the position coordinates of the resampled particles unchanged; Makes a random correction to the search direction of the resampled particles.

10. A road topology construction system, characterized in that: The system comprises: A lane mapping module, at least for mapping the sensed lane lines to a sampling grid map; A topology search module, at least for iteratively searching the area covered by the lane line to identify topological nodes, and recording topological changes and topological boundary information after each search; The topology splicing module is used to connect the topology nodes in sequence according to the recorded topology change order to construct a complete road topology after searching to the end of the lane line.

11. A readable storage medium, characterized in that: The readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the road topology construction method according to any one of claims 1 to 9 is implemented.

Images