A method for generating structured road scenes suitable for intelligent vehicles

Abstract

This disclosure provides a method for generating structured road scenes suitable for intelligent vehicles, comprising: selecting a baseline road scene structure and constructing its "node-edge" graph structure model. The nodes in this model include center nodes corresponding to intersections and end nodes corresponding to entrances and exits of the road scene structure. The edges in this model connect the nodes, corresponding to road segments formed by all adjacent lanes in both directions between the entrances and exits of the road scene structure; randomizing the nodes and edges in the "node-edge" graph structure model to generate a randomized "node-edge" graph structure model; determining whether the randomized "node-edge" graph structure model is compliant; if not, regenerating the randomized "node-edge" graph structure model; and outputting the road scene structure corresponding to the compliant randomized "node-edge" graph structure model. This disclosure can solve problems such as poor generalization performance of driving strategies trained in specific scenarios and poor real-time performance of online path planning in intelligent vehicles.

Description

Technical Field

[0001] This disclosure pertains to the field of intelligent vehicle decision-making and control, and specifically relates to a method, apparatus, and storage medium for generating structured road scenes suitable for intelligent vehicles. Background Technology

[0002] Intelligent vehicles possess significant potential advantages in reducing accident losses, alleviating the energy crisis, and improving traffic efficiency, and are expected to become the primary mode of transportation in the future. With the development of artificial intelligence technologies, such as reinforcement learning, learning-based driving strategies have received widespread attention and application.

[0003] Advanced autonomous driving requires driving strategies to possess high real-time decision-making and control capabilities across common road scenarios. On the one hand, existing driving strategies typically learn interactively for a few specific scenarios, making them prone to overfitting and limiting their generalization performance. Therefore, a rich variety of random road scenarios are urgently needed for training learning strategies. On the other hand, to provide reference paths as input to the driving strategy, existing path planning methods need to simultaneously consider road structure, traffic rules, and information on surrounding dynamic traffic participants to plan feasible paths online, which makes it difficult to guarantee real-time performance in complex scenarios. Summary of the Invention

[0004] This disclosure aims to address one of the aforementioned problems.

[0005] To address these issues, the first aspect of this disclosure provides a structured road scene generation method that can resolve problems such as poor generalization performance of driving strategies trained in specific scenarios and poor real-time performance of online path planning in intelligent vehicles. The method includes:

[0006] 1) Select a benchmark road scene structure and construct a "node-edge" graph structure model of the benchmark road scene structure. The nodes in the "node-edge" graph structure model include the center node corresponding to the intersection and the end node corresponding to the entrance and exit of the road scene structure. The edges in the "node-edge" graph structure model are used to connect the nodes and correspond to the road segments formed by all adjacent lanes in both directions between the entrance and exit of the road scene structure. The size and positional relationship between the nodes and the edges are used to characterize the structural information of the road scene.

[0007] 2) Randomize the nodes and edges in the node-edge graph structure model to generate a randomized node-edge graph structure model;

[0008] 3) Determine whether the randomized "node-edge" graph structure model is compliant based on the set node spacing threshold and edge angle threshold. If it is not compliant, return to step 2) to regenerate the randomized "node-edge" graph structure model; output the road scene structure corresponding to the compliant randomized "node-edge" graph structure model as the first result.

[0009] In some embodiments, a square crossroads is selected as the baseline road scene structure.

[0010] In some embodiments, step 2) specifically includes:

[0011] 2-1) Randomize the edges, including: randomizing the number of edges connected to the central node and defining its upper and lower limits. The lower limit of the number of edges is 2, corresponding to a multi-lane scenario structure; when the number of edges is greater than 2, it corresponds to a multi-branch intersection scenario structure, and the number of edges is equal to the number of intersection branches; randomize the length of the edges connected to the central node and define its upper and lower limits. The lower limit of the edge length should ensure that drivers have sufficient time to adapt to the current road segment and prepare to enter the next road segment; randomize the width of the edges connected to the central node and define its upper and lower limits. The width of the edge corresponds to the total width of all lanes in both directions.

[0012] 2-2) Randomize the nodes, including: randomize the radius of the center node and define its upper and lower limits. The center node is the outer circle of the area where the intersection is located, and given the width of the road segment, the radius of the center node should be greater than the width of the side; randomly perturb the position of each node by adding a bias that follows a normal distribution.

[0013] In some embodiments, the positions of each node are randomly perturbed using at least one of the following two methods:

[0014] The first method: fix the position of any node and randomize the positions of the remaining nodes. Specifically, add biases to the planar coordinates of the remaining nodes, and the magnitude of the biases follows a normal distribution.

[0015] The second method is to fix the position of each central node in sequence and randomize the included angles between the edges connected to it. Specifically, if there are m edges connected to the central nodes, there are m included angles between adjacent edges and the sum is 360°. First, let the size of each of the m angles be equal to 360° / m. Then, add a bias to each of the m angles. The size of the bias follows a normal distribution. Then, normalize the m angles so that their sum is 360°.

[0016] In some embodiments, determining whether the generated randomized "node-edge" graph structure model is compliant specifically includes:

[0017] If the distance between two adjacent nodes in the generated randomized "node-edge" graph structure model is less than or equal to the lower limit of the edge length, or the angle between two adjacent edges is less than or equal to the angle threshold of the edge, then the generated randomized "node-edge" graph structure model is considered non-compliant; if the distance between two adjacent nodes in the generated randomized "node-edge" graph structure model is between the lower and upper limits of the edge length, and the angle between two adjacent edges is between the angle threshold of the edge and 180°, then the generated randomized "node-edge" graph structure model is considered compliant.

[0018] In some embodiments, the structured road scene generation method further includes:

[0019] 4) Generate a static path that considers intersection rule information from the obtained first result, and use this as the second result output. The specific steps are as follows:

[0020] 4-1) Generate a smooth and gentle feasible path from the first result.

[0021] For driving situations where there are lanes and it is necessary to maintain a straight driving position, the center line of the lane is directly taken as the feasible path.

[0022] For driving situations requiring lane changes, intersection crossings, and U-turns, curves are used to connect the current lane and the target lane to form a smooth and gentle feasible path.

[0023] 4-2) Express traffic rule information as different static paths

[0024] Assume an intersection has three traffic light conditions: the traffic light is in the green phase dedicated to the current driving task, the traffic light is in a conflicting green phase where straight and turning traffic are allowed simultaneously, and the traffic light is in the red phase; assume there are four driving tasks: straight, left turn, right turn, and U-turn; the static path representation methods for each driving task under each traffic light condition are as follows:

[0025] When the traffic light is in the green phase specific to the current driving task, vehicles can proceed directly. The static paths corresponding to the four driving tasks of going straight, turning left, turning right, and making a U-turn will adopt the feasible paths described above.

[0026] When the traffic light is in a conflicting green light phase where both straight-ahead and turning vehicles are allowed to proceed simultaneously: For vehicles tasked with going straight, their static path follows the described feasible path; for vehicles tasked with turning left, if their path intersects with an oncoming straight-ahead vehicle at a first intersection, the static path of the left-turning vehicle is planned only up to the first intersection on the feasible path of the left-turning vehicle before being cut off, with a safety margin; for vehicles tasked with turning right, if their path intersects with an oncoming straight-ahead vehicle to their left, the static path of the right-turning vehicle is planned only up to the second intersection on the feasible path of the right-turning vehicle before being cut off, with a safety margin; for vehicles tasked with making a U-turn, if their path intersects with an oncoming straight-ahead vehicle at a third intersection, the static path of the U-turning vehicle is planned only up to the third intersection on the feasible path of the U-turning vehicle before being cut off, with a safety margin.

[0027] When the traffic light is in the red phase: the static paths corresponding to the four tasks of going straight, turning left, turning right, and making a U-turn are only planned up to the stop line at the intersection of the feasible path for each driving task and are then cut off, leaving a parking margin.

[0028] In some embodiments, step 4-1), connecting the current lane and the target lane using a curve to form a smooth and gentle feasible path, includes:

[0029] 4-1-1) Let P1 be the end point of the current lane where the intelligent vehicle is located, and P2 be the starting point of the target lane that the intelligent vehicle needs to enter. Let the coordinates be (x1, y1) and (x2, y2) respectively. Let the orientation angles of the current lane and the target lane be recorded as φ1 and φ2 respectively. Let the orientation angle φ1 of the current lane be defined as the angle between the line connecting P1 and P2 and the center line of the current lane. Let the orientation angle φ2 of the target lane be defined as the angle between the line connecting P1 and P2 and the center line of the target lane.

[0030] 4-1-2) Calculate the distance L between P1 and P2. Starting from P1 and P2 respectively, extend αL into the intersection along their respective lane orientation angles φ1 and φ2 to obtain two auxiliary points P3 and P4, with coordinates denoted as (x3, y3) and (x4, y4) respectively. The specific calculation formula is shown in equation (1), where the hyperparameter α is used to adjust the smoothness of the path. The larger α is, the smaller the average curvature of the path.

[0031]

[0032] 4-1-3) Calculate the parametric equation of the continuous path segment using four points P1, P2, P3, and P4. The specific calculation formula is shown in equation (2). Different values ​​of parameter η correspond to different points P on the continuous path segment. R In particular, when η = 0, the calculated P R=P1; When η=1, calculate P R =P2, discretizing η will achieve the discretization of this continuous path segment:

[0033] P R =P1(1-η) 3 +3P3η(1-η) 2 +3P4η 2 (1-η)+P2η 3 ,η∈[0,1] (2)

[0034] The path determined by formula (2) is the feasible path.

[0035] The structured road scene generation apparatus provided in the second aspect of this disclosure includes:

[0036] The first module is configured to select a baseline road scene structure and construct a "node-edge" graph structure model of the baseline road scene structure. The nodes in the "node-edge" graph structure model include center nodes corresponding to intersections and end nodes corresponding to entrances and exits of the road scene structure. The edges in the "node-edge" graph structure model are used to connect the nodes and correspond to the road segments formed by all adjacent lanes in both directions between the entrances and exits of the road scene structure. The size and positional relationship between the nodes and the edges are used to characterize the structural information of the road scene.

[0037] The second module is configured to randomize the nodes and edges in the node-edge graph structure model to generate a randomized node-edge graph structure model.

[0038] The third module is configured to determine whether the randomized "node-edge" graph structure model is compliant based on the set node spacing threshold and edge angle threshold. If it is not compliant, the randomized "node-edge" graph structure model is regenerated. The road scene structure corresponding to the compliant randomized "node-edge" graph structure model is output as the first result.

[0039] In some embodiments, the structured road scene generation apparatus further includes:

[0040] The fourth module is configured to generate a static path that takes into account intersection rule information from the first result obtained, and use this as the output of the second result.

[0041] A computer-readable storage medium is provided in the third aspect of this disclosure, the computer-readable storage medium storing computer instructions for causing the computer to execute the structured road scene generation method according to any embodiment of the first aspect of this disclosure.

[0042] This disclosure has the following characteristics and beneficial effects:

[0043] This disclosure presents a "node-edge" abstraction method to generate road scene structures. By applying reasonable randomization operations to node positions, node radii, edge angles, and edge geometric dimensions, it generates highly diverse road scene structures for training learning-based driving strategies. Therefore, it addresses the problems of poor generalization performance and poor real-time performance of online path planning in intelligent vehicles trained in specific scenarios. Furthermore, this disclosure embeds road scene structure information, traffic light rules, and yielding rules into the static paths of the scene through the design of general rules. These static paths can be pre-stored in an electronic map, providing alternative feasible paths for intelligent vehicle driving strategies in real time. During training or application, the driving strategy can directly read the corresponding feasible static paths without online planning of reference paths, ensuring real-time decision-making. Attached Figure Description

[0044] Figure 1 This is a flowchart of a method for generating structured road scenes for intelligent vehicles, provided in the first aspect of this disclosure.

[0045] Figure 2 This is a schematic diagram illustrating the method provided in the first aspect of this disclosure, which abstracts the road scene structure as a "node-edge" graph structure.

[0046] Figure 3 (a) to (c) are schematic diagrams of static path planning methods for different road scene structures provided in the first aspect of the present disclosure.

[0047] Figure 4 (a1) to (c4) are static path diagrams for different driving tasks under different traffic light conditions in the method provided in the first aspect of the present disclosure.

[0048] Figure 5 (a) to (h) are the road scene structure and the corresponding static path result diagram randomly generated in the embodiments of this disclosure, respectively.

[0049] Figure 6 This is a schematic diagram of the structure of an electronic device provided in a third aspect embodiment of the present disclosure. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] Conversely, this application covers any substitutions, modifications, equivalent methods, and schemes made within the spirit and scope of this application as defined by the claims. Furthermore, to provide the public with a better understanding of this application, certain specific details are described in detail below. However, those skilled in the art will fully understand this application even without these detailed descriptions.

[0052] The method for generating structured road scenes for intelligent vehicles provided in the first aspect of this disclosure, see also... Figure 1 ,include:

[0053] 1) Select a baseline road scene structure and construct a node-edge graph structure model of the baseline road scene structure. See [link to relevant documentation]. Figure 2 In the node-edge graph structure model, the nodes include the central node corresponding to the intersection and the end nodes corresponding to the entrances and exits of the road scene structure. The edges in the node-edge graph structure model are used to connect the nodes and correspond to the road segments formed by all adjacent lanes in both directions between the entrances and exits of the road scene structure. The size and positional relationship of the nodes and edges can characterize the structural information of the road scene.

[0054] 2) Randomize the nodes and edges in the node-edge graph structure model to generate a randomized node-edge graph structure model;

[0055] 3) Determine whether the generated randomized "node-edge" graph structure model is compliant based on the set node spacing threshold and edge angle threshold. If it is not compliant, return to step 2) to regenerate the randomized "node-edge" graph structure model; and output the road scene structure corresponding to the compliant randomized "node-edge" graph structure model as the first result.

[0056] In some embodiments, a square crossroads is selected as the baseline road scene structure.

[0057] In some embodiments, the nodes and edges in the node-edge graph structure model are randomized to generate a randomized node-edge graph structure model, specifically including:

[0058] 2-1) Randomize the edges

[0059] The number of edges connected to the central node is randomized, and its upper and lower limits are defined: an edge is an abstract representation of a road segment, and the minimum number of edges is 2, corresponding to a multi-lane scenario structure; when the number of edges is greater than 2, it corresponds to a multi-branch intersection scenario structure, and the number of edges is equal to the number of intersection branches. Optionally, the upper and lower limits of the number of edges can be taken as 2 to 8.

[0060] The length of the edges connected to the central node is randomized, and its upper and lower limits are defined: the edge length corresponds to the road segment length. The road segment length should not be too short, requiring sufficient time for the driver to adapt to the current road segment and prepare to enter the next road segment; at the same time, the road segment length should not be too long, otherwise excessively long straight sections will make the road scene structure insufficiently rich. Optionally, the edge length range can be from 50m to 1000m.

[0061] The width of the edges connected to the central node is randomized, and its upper and lower limits are defined: the edge width corresponds to the total width of all lanes in both directions, so the edge width is equal to the number of lanes in both directions multiplied by the lane width. For larger intersections with median strips, the width of the median strip is also included in the calculation. Optionally, the number of lanes in both directions can be 2 to 12, the lane width can be 3m to 3.75m, and the width of the median strip, etc., can be 0m to 10m, then the upper and lower limits of the edge width can be 6m to 55m.

[0062] 2-2) Randomize the nodes

[0063] The radius of the center node is randomized, and its upper and lower limits are defined: The radius of the center node determines the size of the intersection. Specifically, the intersection can be considered as an irregular polygon, and the center node is the circumcircle of this irregular polygon. It should be noted that, given the width of the road segment, the radius of the center node should be greater than the width of the side. Optionally, the upper and lower limits of the node radius can be taken as 20m to 60m.

[0064] Randomly perturbing the positions of each node: Since nodes are connected by edges, perturbing the planar coordinates of each node can change the structure of the entire "node-edge" graph structure model. Because the structure of the graph structure model depends only on the relative positions of the nodes, embodiments of this disclosure employ any one or both of the following methods simultaneously to randomly perturb the positions of the nodes:

[0065] The first method: Fix the position of any node and randomize the positions of the remaining nodes. Specifically, add a bias (Δx, Δy) to the planar coordinates (x, y) of each of the remaining nodes (the magnitude of the bias added to the planar coordinates of the remaining nodes can be the same or different). The magnitude of the bias follows a normal distribution N(μ, σ). 2 The sample is obtained from the sample, where the value of μ is 0 and the value of σ can be from 0 to 10;

[0066] The second method: Fix the position of each central node sequentially, and randomize the angles between the edges connected to it. Specifically, if there are m edges connected to the central nodes, there are m angles between adjacent edges, and their sum is 360°. First, let the size of each of these m angles be equal to 360° / m, and then add a bias Δφ to each (the bias added to the m angles can be the same or different). The bias sizes follow a normal distribution N(μ,σ). 2 The m angles are sampled and obtained, where μ is 0 and σ can be 0 to 10. These m angles are then normalized so that their sum is 360°. In some embodiments, determining whether the generated randomized "node-edge" graph structure model is compliant specifically includes:

[0067] If the distance between two adjacent nodes (i.e., the length of the edge connecting the two nodes) in the generated randomized node-edge graph structure model is less than or equal to the lower limit of the edge length, or the angle between two adjacent edges is less than or equal to the edge angle threshold (e.g., 30°), then the generated randomized node-edge graph structure model is considered non-compliant. If the distance between two adjacent nodes (i.e., the length of the edge connecting the two nodes) in the generated randomized node-edge graph structure model is between the lower and upper limits of the edge length, and the angle between two adjacent edges is between the edge angle threshold and 180°, then the generated randomized node-edge graph structure model is considered compliant.

[0068] The structured road scene generation method for intelligent vehicles provided in the first aspect of this disclosure represents the road scene structure through the abstraction of "node-edge". It can generate new road scene structures by applying reasonable randomization operations on aspects such as node position, node radius, angle between edges, and edge size. It has the advantage of being able to generate sufficiently diverse road scene structures for training learning strategies.

[0069] In some embodiments, the structured road scene generation method for intelligent vehicles provided by the first aspect of this disclosure further includes: 4) generating a static path that considers intersection rule information from the obtained first result, and using this as the output of the second result, the specific steps of which are as follows:

[0070] 4-1) Generate smooth and gentle feasible paths in the first result.

[0071] For driving situations where there are lanes and it is necessary to keep the lanes straight, the center line of the lane is taken directly as the feasible path.

[0072] For driving scenarios requiring lane changes, intersection crossings, or U-turns, a curve connects the current lane and the target lane to create a smooth and feasible path. The specific formula and steps for generating the connecting curve are explained below:

[0073] 4-1-1) Let P1 be the end point of the current lane where the intelligent vehicle is located (the end point is the intersection of the center line of the current lane where the intelligent vehicle is located and the vehicle stop line of the intersection), and P2 be the starting point of the target lane that the intelligent vehicle needs to enter (the starting point is the intersection of the center line of the target lane that the intelligent vehicle needs to enter and the vehicle stop line of the intersection). Let the coordinates be (x1, y1) and (x2, y2) respectively. Let the orientation angles of the current lane and the target lane be recorded as φ1 and φ2 respectively. Let the orientation angle φ1 of the current lane be defined as the angle between the line connecting P1 and P2 and the center line of the current lane. Let the orientation angle φ2 of the target lane be defined as the angle between the line connecting P1 and P2 and the center line of the target lane.

[0074] 4-1-2) Calculate the distance L between P1 and P2. Starting from P1 and P2 respectively, extend αL into the intersection along their respective lane orientation angles φ1 and φ2 to obtain two auxiliary points P3 and P4, with coordinates denoted as (x3, y3) and (x4, y4) respectively. The specific calculation formula is shown in equation (1), where the hyperparameter α∈[0.2, 0.8] can adjust the smoothness of the path. The larger α is, the smaller the average curvature of the path, that is, the smoother the generated static path.

[0075]

[0076] 4-1-3) Calculate the parametric equation of the continuous path segment using four points P1, P2, P3, and P4. The specific calculation formula is shown in equation (2). Different values ​​of parameter η correspond to different points P on the continuous path segment. R In particular, when η = 0, the calculated P R =P1; When η=1, calculate P R =P2, discretizing η will achieve the discretization of this continuous path segment.

[0077] P R =P1(1-η) 3 +3P3η(1-η) 2 +3P4η 2 (1-η)+P2η 3 ,η∈[0,1] (2)

[0078] The parameters of the generated path include the coordinates of four control points (x1,y1), (x2,y2), (x3,y3), and (x4,y4), the orientations of the current lane and the target lane φ1 and φ2, and the hyperparameter α. Saving these feature parameters enables the storage of this static path segment. Figure 3 Images (a) through (c) show the static path planning results of the method disclosed herein for three driving tasks: lane change, left turn, and U-turn.

[0079] 4-2) Express traffic rule information as different static paths

[0080] The feasible paths generated in step 4-1) only consider the static information of the road structure. However, in real road scenarios, traffic light phase schemes and yield rules are also static information that does not change due to dynamic traffic participants. Therefore, the method disclosed in this paper expresses the static information of traffic rules as different static paths for the intelligent vehicle decision-making and control system to retrieve in real time. Specifically, such as Figure 4 As shown, taking a crossroads scenario as an example, using a two-way six-lane road, there are three traffic light conditions: the traffic light is in the green phase dedicated to the current driving task (i.e., for the traffic flow of this driving task, there are no other traffic flows conflicting with it in this phase), the traffic light is in a conflicting green phase where straight and turning traffic are allowed simultaneously, and the traffic light is in a red phase; there are four driving tasks: straight, left turn, right turn, and U-turn; the static path representation method for each driving task under each traffic light condition will be explained below:

[0081] The first traffic light condition is during the green light phase specific to the current driving task. In this case, vehicles can proceed directly. Therefore, the static paths corresponding to the four tasks—going straight, turning left, turning right, and making a U-turn—are consistent with the feasible paths generated in step 4-1). (See [link]). Figure 4 The line segments shown by the dashed lines in (a1) to (a4);

[0082] The second traffic light condition is a conflicting green light phase where straight-ahead and turning vehicles are allowed to proceed simultaneously. According to traffic rules: in this situation, turning vehicles should yield to straight-ahead vehicles, right-turning vehicles should yield to left-turning vehicles, and U-turning vehicles should yield to turning vehicles. Specifically, for vehicles tasked with going straight, since they have the highest right-of-way priority, their static path is consistent with the feasible path generated in step 4-1), see [link to relevant documentation]. Figure 4 The dashed line segment in (b1) indicates that for vehicles performing left turns, their path intersects with the oncoming straight-going vehicles at a first intersection point. Their static path is planned only up to this first intersection point and then truncated, with a safety margin δ. safe The first intersection point is obtained by extending the feasible path of the left-turning vehicle obtained in step 4-1) along the direction of the starting point of the feasible path by a safety margin to obtain the first intermediate point. Then, the static path of the left-turning vehicle is the path from its starting point to the first intermediate point on the feasible path of the left-turning driving task. See [link to relevant documentation]. Figure 4 The dashed line segment in (b2) indicates that for vehicles performing right turns, their path intersects with the straight-going vehicles to their left at a second intersection point. Similarly, their static path is only planned up to this second intersection point before being cut off, with a safety margin δ. safeThe second intersection point is obtained by extending the feasible path of the right-turning vehicle obtained in step 4-1) along the direction of the starting point of the feasible path by a safety margin to obtain the second intermediate point. The static path of the right-turning vehicle is then the path from its starting point to the second intermediate point on the feasible path of the right-turning vehicle. (See [reference]). Figure 4 The dashed line segment in (b3) indicates that for vehicles performing U-turns, there is a third intersection point between their path and the opposite straight-going vehicles. Their static path is only planned up to this third intersection point and then cut off, leaving a safety margin δ. safe The third intersection point is obtained by extending the feasible path of the U-turn vehicle obtained in step 4-1) along the direction of the starting point of the feasible path by a safety margin to obtain the third intermediate point. Then, the static path of the U-turn vehicle is the path from its starting point to the third intermediate point on the feasible path of the U-turn vehicle. See [link to relevant documentation]. Figure 4 The dashed line segment in (b4) represents the safety margin δ. safe The value can be taken as 3 to 5m to avoid collision, δ safe The larger the value, the greater the allowable space left for avoiding vehicles, that is, the more conservative the approach.

[0083] The third traffic light condition is when the light is red and vehicles are prohibited from entering the intersection. In this case, the static paths for the four tasks of going straight, turning left, turning right, and making a U-turn are only planned up to the stop line at the intersection of the feasible paths for each driving task, and a parking margin δ is left. stop See Figure 4 The line segments shown by the dashed lines in (c1) to (c4). δ stop The value should be slightly greater than half the vehicle length to prevent the front of the vehicle from crossing the stop line; parking margin δ stop The value can be between 2m and 3m.

[0084] Finally, in order to better understand this application, Figure 5 The diagram shows the road scene structure randomly generated by the method of this disclosure and the static path planning results under the corresponding green light phase for the driving task. (a) to (e) are four-way intersections, and (f) to (h) are three-way intersections.

[0085] The structured road scene generation method for intelligent vehicles provided in the first aspect of this disclosure expresses traffic light information, yielding rule information, etc., as different static paths for intelligent vehicles to retrieve in real time, thereby alleviating the computational burden of the decision-making system and improving real-time performance. During the training or application of driving strategies, the corresponding feasible static paths can be directly read without online planning of reference paths, offering the advantage of ensuring real-time decision-making.

[0086] The structured road scene generation apparatus for intelligent vehicles provided in the second aspect of this disclosure includes:

[0087] The first module is configured to select a baseline road scene structure and construct a "node-edge" graph structure model of the baseline road scene structure. The nodes in the "node-edge" graph structure model include the center node corresponding to the intersection and the end nodes corresponding to the entrances and exits of the road scene structure. The edges in the "node-edge" graph structure model are used to connect the nodes and correspond to the road segments formed by all adjacent lanes in both directions between the entrances and exits of the road scene structure. The size and positional relationship of the nodes and edges can characterize the structural information of the road scene.

[0088] The second module is configured to randomize the nodes and edges in the node-edge graph structure model to generate a randomized node-edge graph structure model.

[0089] The third module is configured to determine whether the generated randomized "node-edge" graph structure model is compliant based on the set node spacing threshold and edge angle threshold. If it is not compliant, the randomized "node-edge" graph structure model is regenerated. Finally, the road scene structure corresponding to the compliant randomized "node-edge" graph structure model is output as the first result.

[0090] The structured road scene generation apparatus for intelligent vehicles provided in the second aspect embodiment of this disclosure further includes:

[0091] The fourth module is configured to generate a static path that takes into account intersection rule information from the first result, and use this as the output of the second result.

[0092] To implement the above embodiments, this disclosure also proposes a computer-readable storage medium storing a computer program that is executed by a processor to perform the structured road scene generation method for intelligent vehicles described above.

[0093] The following is for reference. Figure 6 The diagram illustrates a structural schematic of an electronic device suitable for implementing embodiments of the present disclosure. It should be noted that the electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs, desktop computers, and servers. Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0094] like Figure 6As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 102 or a program loaded from a storage device 108 into a random access memory (RAM) 103. The RAM 103 also stores various programs and data required for the operation of the electronic device 100. The processing unit 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.

[0095] Typically, the following devices can be connected to I / O interface 105: input devices 106 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 107 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 108 including, for example, magnetic tapes, hard disks, etc.; and communication devices 109. Communication device 109 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 2 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have instead.

[0096] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, this embodiment includes a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network via communication device 109, or installed from storage device 108, or installed from ROM 102. When the computer program is executed by processing device 101, it performs the functions defined above in the methods of embodiments of this disclosure.

[0097] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0098] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0099] The aforementioned computer-readable medium carries one or more programs. When the aforementioned one or more programs are executed by the electronic device, the electronic device causes the following actions: Selecting a baseline road scene structure, constructing a "node-edge" graph structure model of the baseline road scene structure, wherein the nodes in the "node-edge" graph structure model include center nodes corresponding to intersections and end nodes corresponding to entrances and exits of the road scene structure, and the edges in the "node-edge" graph structure model are used to connect nodes, corresponding to road segments formed by all adjacent lanes in both directions between entrances and exits of the road scene structure, and the size and positional relationship of nodes and edges can characterize the structural information of the road scene; randomizing the nodes and edges in the "node-edge" graph structure model to generate a randomized "node-edge" graph structure model; judging whether the generated randomized "node-edge" graph structure model is compliant based on the set node spacing threshold and edge angle threshold, and if it is not compliant, regenerating the randomized "node-edge" graph structure model; and finally outputting the road scene structure corresponding to the compliant randomized "node-edge" graph structure model as the first result.

[0100] Furthermore, the electronic device also generates a static path that takes into account intersection rule information from the obtained first result, and outputs it as a second result.

[0101] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and Python, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0102] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0103] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0104] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the function involved, as will be understood by those skilled in the art to which embodiments of this application pertain.

[0105] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0106] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0107] Those skilled in the art will understand that implementing all or part of the steps of the methods in the above embodiments can be accomplished by instructing related hardware through a program. The developed program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0108] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0109] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for generating structured road scenes suitable for intelligent vehicles, characterized in that, include: 1) Select a benchmark road scene structure and construct a "node-edge" graph structure model of the benchmark road scene structure. The nodes in the "node-edge" graph structure model include the center node corresponding to the intersection and the end node corresponding to the entrance and exit of the road scene structure. The edges in the "node-edge" graph structure model are used to connect the nodes and correspond to the road segments formed by all adjacent lanes in both directions between the entrance and exit of the road scene structure. The size and positional relationship between the nodes and the edges are used to characterize the structural information of the road scene. 2) Randomize the nodes and edges in the "node-edge" graph structure model to generate a randomized "node-edge" graph structure model. Randomization of nodes includes: randomizing the radius of the center node and defining its upper and lower limits; the center node is the circumcircle of the area where the intersection is located, and given the width of the road segment, the radius of the center node should be greater than the width of the edge; randomly perturbing the position of each node by adding a bias that follows a normal distribution; randomizing edges includes: randomizing the number of edges connected to the center node and defining its upper and lower limits; the lower limit of the number of edges is... This corresponds to a multi-lane scenario structure; the number of edges is greater than... When the intersection is multi-branched, the number of edges equals the number of bifurcations. The length of the edge connected to the central node is randomized, and its upper and lower limits are defined. The lower limit of the edge length should ensure that the driver has sufficient time to adapt to the current road segment and prepare to enter the next road segment. The width of the edge connected to the central node is randomized, and its upper and lower limits are defined. The width of the edge corresponds to the total width of all lanes in both directions. The positions of each node are randomly perturbed using at least one of the following two methods: The first method: fix the position of any node and randomize the positions of the remaining nodes. Specifically, add biases to the planar coordinates of the remaining nodes, and the magnitude of the biases follows a normal distribution. The second method: Fix the position of each central node sequentially, and randomize the included angles between the edges connected to it. Specifically, let... If an edge is connected to the center node, then there are a total of [number] edges between adjacent edges. The included angles are summed as... ,shilling The size of each angle is equal to And then Add a bias to each corner, the magnitude of the bias following a normal distribution, and then... Normalizing the angles makes their sum equal to... ; 3) Determine whether the randomized "node-edge" graph structure model is compliant based on the set node spacing threshold and edge angle threshold. If it is not compliant, return to step 2) to regenerate the randomized "node-edge" graph structure model; output the road scene structure corresponding to the compliant randomized "node-edge" graph structure model as the first result.

2. The structured road scene generation method according to claim 1, characterized in that, A square intersection was selected as the baseline road scene structure.

3. The structured road scene generation method according to claim 1, characterized in that, Determining whether the generated randomized "node-edge" graph structure model is compliant specifically includes: If the distance between two adjacent nodes in the generated randomized "node-edge" graph structure model is less than or equal to the lower limit of the edge length, or the angle between two adjacent edges is less than or equal to the angle threshold of the edge, then the generated randomized "node-edge" graph structure model is considered non-compliant. If the distance between two adjacent nodes in the generated randomized "node-edge" graph structure model is between the lower and upper limits of the edge length, and the angle between two adjacent edges is between the angle threshold of the edge and the upper limit of the edge length, then the generated randomized "node-edge" graph structure model is considered non-compliant. If the generated randomized "node-edge" graph structure model is between these two, it is considered compliant.

4. The method for generating structured road scenes according to any one of claims 1 to 3, characterized in that, Also includes: 4) Generate a static path that considers intersection rule information from the obtained first result, and use this as the second result output. The specific steps are as follows: 4-1) Generate a smooth and gentle feasible path from the first result. For driving situations where there are lanes and it is necessary to maintain a straight driving position, the center line of the lane is directly taken as the feasible path. For driving situations requiring lane changes, intersection crossings, and U-turns, curves are used to connect the current lane and the target lane to form a smooth and gentle feasible path. 4-2) Express traffic rule information as different static paths Assume an intersection has three traffic light conditions: the traffic light is in the green phase dedicated to the current driving task, the traffic light is in a conflicting green phase where straight and turning traffic are allowed simultaneously, and the traffic light is in the red phase; assume there are four driving tasks: straight, left turn, right turn, and U-turn; the static path representation methods for each driving task under each traffic light condition are as follows: When the traffic light is in the green phase specific to the current driving task, vehicles can proceed directly. The static paths corresponding to the four driving tasks of going straight, turning left, turning right, and making a U-turn will adopt the feasible paths described above. When the traffic light is in a conflicting green light phase where both straight-ahead and turning vehicles are allowed to proceed simultaneously: For vehicles tasked with going straight, their static path follows the described feasible path; for vehicles tasked with turning left, if their path intersects with an oncoming straight-ahead vehicle at a first intersection, the static path of the left-turning vehicle is planned only up to the first intersection on the feasible path of the left-turning vehicle before being cut off, with a safety margin; for vehicles tasked with turning right, if their path intersects with an oncoming straight-ahead vehicle to their left, the static path of the right-turning vehicle is planned only up to the second intersection on the feasible path of the right-turning vehicle before being cut off, with a safety margin; for vehicles tasked with making a U-turn, if their path intersects with an oncoming straight-ahead vehicle at a third intersection, the static path of the U-turning vehicle is planned only up to the third intersection on the feasible path of the U-turning vehicle before being cut off, with a safety margin. When the traffic light is in the red phase: the static paths corresponding to the four tasks of going straight, turning left, turning right, and making a U-turn are only planned up to the stop line at the intersection of the feasible path for each driving task and are then cut off, leaving a parking margin.

5. The structured road scene generation method according to claim 4, characterized in that, In step 4-1), the current lane and the target lane are connected by a curve to form a smooth and gentle feasible path, including: 4-1-1) Take the end point of the current lane where the intelligent vehicle is located as... The starting point of the target lane that the intelligent vehicle needs to enter is The coordinates are denoted as follows: and And record the orientation angles of the current lane and the target lane, denoted as . and Define the orientation angle of the current lane. for and The angle between the line connecting the two lanes and the current lane centerline defines the orientation angle of the target lane. for and The angle between the line connecting the two lanes and the center line of the target lane; 4-1-2) Calculation and Distance between From respectively and Depart, heading towards the corner along their respective lanes. and Extending into the intersection Distance, obtained and Two auxiliary points, their coordinates are denoted as follows: and The specific calculation formula is shown in the formula. hyperparameters Used to adjust the smoothness of the path. The larger the value, the smaller the average curvature of the path. 4-1-3) by Calculate the parametric equations of this continuous path segment using four points. The specific calculation formula is shown in the equation. ,parameter Different values ​​correspond to different points on the continuous path of this segment. In particular, when At that time, the calculation ;when At that time, the calculation ,right Discretization can be performed to discretize this continuous path segment: The path determined by formula (2) is the feasible path.

6. A structured road scene generation device suitable for intelligent vehicles, characterized in that, include: The first module is configured to select a baseline road scene structure and construct a "node-edge" graph structure model of the baseline road scene structure. The nodes in the "node-edge" graph structure model include center nodes corresponding to intersections and end nodes corresponding to entrances and exits of the road scene structure. The edges in the "node-edge" graph structure model are used to connect the nodes and correspond to the road segments formed by all adjacent lanes in both directions between the entrances and exits of the road scene structure. The size and positional relationship between the nodes and the edges are used to characterize the structural information of the road scene. The second module is configured to randomize the nodes and edges in the "node-edge" graph structure model to generate a randomized "node-edge" graph structure model. Randomizing nodes includes: randomizing the radius of the center node and defining its upper and lower limits; the center node is the circumcircle of the area where the intersection is located, and given the width of the road segment, the radius of the center node should be greater than the width of the edge; and randomly perturbing the position of each node by adding a bias that follows a normal distribution. Randomizing edges includes: randomizing the number of edges connected to the center node and defining its upper and lower limits; the lower limit of the number of edges is... This corresponds to a multi-lane scenario structure; the number of edges is greater than... When the intersection is multi-branched, the number of edges equals the number of bifurcations. The length of the edge connected to the central node is randomized, and its upper and lower limits are defined. The lower limit of the edge length should ensure that the driver has sufficient time to adapt to the current road segment and prepare to enter the next road segment. The width of the edge connected to the central node is randomized, and its upper and lower limits are defined. The width of the edge corresponds to the total width of all lanes in both directions. The positions of each node are randomly perturbed using at least one of the following two methods: The first method: fix the position of any node and randomize the positions of the remaining nodes. Specifically, add biases to the planar coordinates of the remaining nodes, and the magnitude of the biases follows a normal distribution. The second method: Fix the position of each central node sequentially, and randomize the included angles between the edges connected to it. Specifically, let... If an edge is connected to the center node, then there are a total of [number] edges between adjacent edges. The included angles are summed as... ,shilling The size of each angle is equal to And then Add a bias to each corner, the magnitude of the bias following a normal distribution, and then... Normalizing the angles makes their sum equal to... ; The third module is configured to determine whether the randomized "node-edge" graph structure model is compliant based on the set node spacing threshold and edge angle threshold. If it is not compliant, the randomized "node-edge" graph structure model is regenerated. The road scene structure corresponding to the compliant randomized "node-edge" graph structure model is output as the first result.

7. The structured road scene generation device according to claim 6, characterized in that, Also includes: The fourth module is configured to generate a static path that takes into account intersection rule information from the first result obtained, and use this as the output of the second result.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the structured road scene generation method for intelligent vehicles as described in any one of claims 1 to 5.

Images