Automatic driving simulation scene generation method and device, equipment, medium and product

Abstract

The invention relates to an automatic driving simulation scene generation method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: performing path planning of a simulation scene based on multi-source data to obtain a target planning path; determining at least one passing intersection in the target planning path, screening traffic flow segment data according to respective intersection traffic flow data of the passing intersections in the multi-source data, and determining intersection traffic flow segment data meeting an automatic driving simulation scene generation condition; for the adjacent passing intersections, according to the intersection traffic flow data of the adjacent passing intersections, performing traffic flow data reasoning reduction on the connection road sections with data missing between the adjacent passing intersections to obtain road section traffic flow reduction data; and generating an automatic driving simulation scene according to the intersection traffic flow data, the screened intersection traffic flow fragment data and the road section traffic flow reduction data. By adopting the method, the automatic driving simulation test effect can be improved.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for generating an autonomous driving simulation scene. Background Art

[0002] With the accelerated development of global intelligent connected vehicle technology, autonomous driving simulation testing has become a necessary means to verify the safety, stability and performance of autonomous driving systems, but autonomous driving simulation testing must rely on collected autonomous driving simulation scenarios.

[0003] Generally, autonomous driving simulation scenarios can be generated with the help of data collected by autonomous driving vehicles during the driving process for use in autonomous driving simulation tests. However, the autonomous driving simulation scenarios generated in this way usually find it difficult to cover diverse traffic conditions, thus affecting the results of autonomous driving simulation tests. Summary of the invention

[0004] Based on this, in response to the above-mentioned technical problems, the present application at least provides a method, device, computer equipment, computer-readable storage medium and computer program product for generating autonomous driving simulation scenarios that are conducive to improving the effect of autonomous driving simulation testing.

[0005] In a first aspect, the present application provides a method for generating an autonomous driving simulation scene, comprising: obtaining multi-source data for autonomous driving simulation testing, performing path planning for the simulation scene based on the multi-source data, and obtaining a target planned path, wherein the multi-source data includes traffic flow data and map data; determining at least one intersection in the target planned path, and screening traffic flow fragment data based on the intersection traffic flow data of at least one intersection in the multi-source data, and determining intersection traffic flow fragment data that meets the conditions for generating an autonomous driving simulation scene; for every two adjacent intersections in at least one intersection, performing traffic flow data inference and restoration for a connecting section where data is missing between adjacent intersections based on the intersection traffic flow data of the adjacent intersections, and obtaining section traffic flow restoration data for the connecting section; generating an autonomous driving simulation scene based on the intersection traffic flow data, the intersection traffic flow fragment data that meets the conditions for generating an autonomous driving simulation scene, and the section traffic flow restoration data.

[0006] The above-mentioned method for generating a simulation scene for autonomous driving, after performing path planning for the simulation scene based on multi-source data (including traffic flow data and map data) and obtaining the target planned path, first determines at least one intersection in the target planned path, and then screens the traffic flow segment data based on the intersection traffic flow data of at least one intersection in the multi-source data, and determines the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene, so as to assist in generating the autonomous driving simulation scene. Furthermore, for each two adjacent intersections in at least one intersection, the traffic flow data can be inferred and restored for the connecting road section with missing data between the adjacent intersections based on the intersection traffic flow data of the adjacent intersections, so as to complete the traffic flow data on the target path and obtain the section traffic flow restoration data of the connecting road section. Based on this, on the basis of the original intersection traffic flow data, combined with the intersection traffic flow fragment data that meets the conditions for generating autonomous driving simulation scenarios, and the road section traffic flow restoration data obtained by reasoning and restoration, the traffic scene on the target planning path is made clearer, continuous and reasonable, which is conducive to generating effective autonomous driving simulation scenarios based on the above traffic flow data, and thus improving the effect of autonomous driving simulation testing based on autonomous driving simulation scenarios.

[0007] In one of the embodiments, path planning of a simulation scenario is performed based on multi-source data to obtain a target planned path, including: screening traversable areas based on map data in the multi-source data to obtain a topological map including traversable areas; obtaining path planning constraints of the simulation scenario, performing path planning according to the path planning constraints and the topological map to obtain at least one sub-planning path arranged in sequence; and connecting at least one sub-planning path in sequence to obtain a target planned path.

[0008] In this embodiment, path planning can be used to obtain a target planning path formed by connecting at least one sub-planning path, so that the target planning path covers as rich and complete traffic scenes as possible, which is conducive to the subsequent screening, reasoning and restoration of traffic flow data on the target planning path, making the traffic scene on the target planning path clearer, continuous and reasonable, thereby generating an effective autonomous driving simulation scene and improving the effect of autonomous driving simulation testing based on the autonomous driving simulation scene.

[0009] In a specific example, the path planning constraints of the simulation scenario are obtained, including: constructing a path coverage evaluation function, and maximizing the value of the path coverage evaluation function as the first constraint; the value of the path coverage evaluation function represents the degree of diversity of sub-planning paths contained in a path; sequentially adjacent sub-planning paths satisfying the end-to-end connection relationship is used as the second constraint, and the total mileage of at least one sub-planning path reaches the expected mileage as the third constraint; the first constraint, the second constraint and the third constraint are combined to obtain the path planning constraints.

[0010] In this embodiment, long-term continuous path planning can be reasonably performed through multiple constraints, so as to obtain a target planning path that covers as rich and complete traffic scenarios as possible.

[0011] In some embodiments, traffic flow segment data is screened based on the intersection traffic flow data of at least one passing intersection in multi-source data to determine the intersection traffic flow segment data that meets the conditions for generating an autonomous driving simulation scene, including: obtaining the intersection traffic flow data of at least one passing intersection from the multi-source data, the intersection traffic flow data including: intersection traffic flow segment data corresponding to at least one traffic scene; for the intersection traffic flow data of each passing intersection, performing traffic scene complexity analysis and traffic scene connectivity analysis on at least one intersection traffic flow segment data in the intersection traffic flow data, respectively, to obtain a scene evaluation result of at least one intersection traffic flow segment data; based on the scene evaluation result of at least one intersection traffic flow segment data, determining the intersection traffic flow segment data that meets the conditions for generating an autonomous driving simulation scene from at least one intersection traffic flow segment data.

[0012] In this embodiment, traffic scene complexity analysis and traffic scene connectivity analysis can be performed on the intersection traffic flow data passing through the target planning path, and intersection traffic flow fragment data that covers both complex traffic scenes and has traffic scene connectivity can be screened out. The screened intersection traffic flow fragment data can then be combined to generate an autonomous driving simulation scene covering rich and complex traffic scenes, thereby improving the autonomous driving simulation test effect.

[0013] In an exemplary embodiment, based on the intersection traffic flow data of each adjacent passing intersection, the traffic flow data of the connecting section with missing data between the adjacent passing intersections is inferred and restored to obtain the section traffic flow restoration data of the connecting section, including: based on the intersection traffic flow data of each adjacent passing intersection, the data flow integrity test is performed on the connecting section between the adjacent passing intersections to obtain the test result; when the test result indicates that there is missing data in the connecting section, the movement trajectory of the traffic participants in the connecting section is inferred and restored based on the intersection traffic flow data of each adjacent passing intersection to obtain the section traffic flow restoration data of the connecting section.

[0014] In this embodiment, for connecting sections where data is missing between adjacent intersections, the motion trajectories of traffic participants in the connecting sections can be inferred and restored to complete the traffic flow data on the target path, making the traffic scene on the target planned path more continuous and reasonable, which is conducive to the subsequent combination of the section traffic flow restoration data obtained after reasoning and restoration to generate an autonomous driving simulation scene covering rich and complex traffic scenes, thereby improving the effect of autonomous driving simulation testing.

[0015] In a possible implementation, the motion trajectories of traffic participants in the connecting section are inferred and restored based on the intersection traffic flow data of each adjacent passing intersection to obtain the section traffic flow restoration data of the connecting section, including: determining the first intersection and the second intersection in the adjacent passing intersections, as well as the first traffic participant object that disappears at the first intersection and the second traffic participant object that appears at the second intersection based on the intersection traffic flow data of each adjacent passing intersection; the first intersection is connected to the second intersection via the connecting section; the first traffic participant object and the second traffic participant object are paired to obtain at least one matching pair; based on the intersection traffic flow data of each of the first intersection and the second intersection, the motion trajectories of the traffic participants in at least one matching pair are inferred and restored to obtain the section traffic flow restoration data of the connecting section.

[0016] In this embodiment, for each adjacent passing intersection, the motion trajectory of the traffic participant objects that are successfully paired in the adjacent passing intersections can be inferred and restored, so that the motion trajectory of the traffic participant objects between the adjacent passing intersections is more reasonable and continuous, thereby making the traffic scenes on the connecting sections between the adjacent passing intersections richer, more reasonable and continuous, which is conducive to the subsequent combination of the road section traffic flow restoration data obtained after reasoning and restoration, to generate an autonomous driving simulation scene covering rich and complex traffic scenes, thereby improving the effect of autonomous driving simulation testing.

[0017] In a second aspect, the present application also provides an autonomous driving simulation scene generation device, including: a target path planning module, which is used to obtain multi-source data for autonomous driving simulation testing, perform path planning for the simulation scene based on the multi-source data, and obtain a target planned path, wherein the multi-source data includes traffic flow data and map data; an intersection traffic flow data screening module, which is used to determine at least one passing intersection in the target planned path, and perform traffic flow segment data screening based on the intersection traffic flow data of each of at least one passing intersection in the multi-source data, to determine the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions; a section traffic flow data restoration module, which is used to perform traffic flow data inference and restoration for a connecting section with missing data between adjacent passing intersections based on the intersection traffic flow data of each of the adjacent passing intersections for every two adjacent passing intersections in at least one passing intersection, to obtain section traffic flow restoration data of the connecting section; a simulation scene generation module, which is used to generate an autonomous driving simulation scene based on the intersection traffic flow data, the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions, and the section traffic flow restoration data.

[0018] In a third aspect, the present application also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method for generating autonomous driving simulation scenarios when executing the computer program.

[0019] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned autonomous driving simulation scene generation method is implemented.

[0020] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the above-mentioned autonomous driving simulation scene generation method when executed by a processor.

[0021] The above-mentioned autonomous driving simulation scene generation device, computer equipment, computer-readable storage medium and computer program product can be based on the original intersection traffic flow data, combined with the selected intersection traffic flow fragment data that meets the autonomous driving simulation scene generation conditions, and the road section traffic flow restoration data obtained by reasoning and restoration, so that the traffic scene on the target planning path is clearer, continuous and reasonable, which is conducive to generating effective autonomous driving simulation scenes based on the above traffic flow data, and thus improving the effect of autonomous driving simulation testing based on autonomous driving simulation scenes.

[0022] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and are used together with the specification to illustrate the technical solution of the present application.

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 An application environment diagram of an autonomous driving simulation scenario generation method in one embodiment;

[0026] Figure 2 A schematic diagram of a flow chart of a method for generating an autonomous driving simulation scene in one embodiment;

[0027] Figure 3 A schematic diagram of connecting two rounds of path planning solutions to obtain a target planning path in an embodiment;

[0028] Figure 4 A schematic diagram of a motion trajectory obtained by inferring and restoring traffic flow data in one embodiment;

[0029] Figure 5 It is a flowchart of a method for generating an autonomous driving simulation scene in another embodiment;

[0030] Figure 6 is a structural block diagram of an automatic driving simulation scene generation device in one embodiment;

[0031] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0033] The method for generating an autonomous driving simulation scene provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Specifically, the server 102 where the autonomous driving simulation platform is located can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The server 102 can communicate with the cloud control platform 104 and the map system 106 through the network, import the intersection traffic flow data of each intersection from the cloud control platform 104 in real time / periodically, and import high-precision map data from the map system 106 in real time / periodically, and store it in the data background for subsequent use. The cloud control platform 104 is interconnected with the collection equipment deployed on the road side, and can store a large amount of intersection traffic flow data collected by the collection equipment for each intersection. The collection equipment can specifically be: various devices that can collect data related to objects on the road, including but not limited to sensors, cameras, radars, etc. installed on the road side.

[0034] Among them, the roadside data (intersection traffic flow data) collected by the collection device, as an important supplement to the data source, is convenient for subsequent continuous and long-mileage autonomous driving simulation tests, and is conducive to reproducing real continuous driving scenes and long-tail scenes that are difficult to find according to predetermined rules, and improving the effectiveness and reliability of subsequent autonomous driving simulation tests based on the generated autonomous driving simulation scenes. The roadside data collected by the collection device can usually comprehensively and completely collect vehicle information or pedestrian information within 50 to 100 meters in all directions of the intersection, but there may be data gaps between intersections (that is, within the road section). In order to make full use of the roadside data for autonomous driving simulation tests, it is necessary to mine and restore continuous traffic scenes for multiple intersections based on the autonomous driving simulation scene generation method provided in this embodiment, so as to discover valuable scenes based on the roadside data of real vehicle-road collaboration and build a continuous simulation scene library.

[0035] Among them, intersection traffic flow data, also known as MEC (Multi-access Edge Computing) data, is data accessed from multiple collection devices. MEC is a distributed computing architecture that can move computing power from the central cloud to the edge of the network (such as roadside collection devices, base stations, etc.), which can help obtain real-time computing services and data processing in a low-latency environment. Map data can specifically include: HD-MAP (High Definition Map) data and V2X-MAP (map based on Vehicle To Everything technology) data. HD-MAP is a high-precision map provided for autonomous driving and advanced driver assistance systems. Unlike traditional navigation maps, it has centimeter-level accuracy and can contain rich road details, such as lane center lines, curbs, traffic signs, traffic lights, slopes and curves. Vehicle To Everything is a communication technology used for intelligent transportation and autonomous driving, which enables vehicles to exchange information with various entities in the surrounding environment in real time to enhance driving safety, efficiency and traffic management. V2X-MAP can provide road environment information such as lane information, traffic signs, and traffic signals to enhance the perception and decision-making capabilities of autonomous vehicles.

[0036] Exemplarily, the server 102 may obtain multi-source data for the autonomous driving simulation test from the data background, and perform path planning of the simulation scene based on the multi-source data to obtain the target planning path, wherein the multi-source data includes traffic flow data and map data. Further, the server 102 may first determine at least one passing intersection in the target planning path of the simulation scene according to the map data, and screen the traffic flow segment data according to the intersection traffic flow data of at least one passing intersection in the multi-source data, and determine the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions. Further, each two adjacent passing intersections in at least one passing intersection are determined according to the map data, and for each two adjacent passing intersections in at least one passing intersection, the server 102 may perform traffic flow data inference and restoration for the connecting road sections with missing data between the adjacent passing intersections according to the intersection traffic flow data of the adjacent passing intersections, and obtain the road section traffic flow restoration data of the connecting road sections, and then generate the autonomous driving simulation scene according to the intersection traffic flow data, the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions, and the road section traffic flow restoration data.

[0037] It should be noted that the operation and maintenance personnel of the autonomous driving simulation platform can also monitor the generation process of the autonomous driving simulation scene through the autonomous driving simulation platform, and can make manual adjustments in time when abnormal situations occur to solve the problems of missing scene data, incoherent or inaccurate scenes that may occur during the generation process, and ensure the accuracy, coherence, authenticity and integrity of the autonomous driving simulation scene. By monitoring and adjusting the scene generation process, the operation and maintenance personnel can effectively avoid logical errors and data deviations in the autonomous driving simulation scene, and ensure that the continuous autonomous driving simulation scene that can accurately simulate the real traffic environment is finally generated, thereby improving the reliability and coverage of the autonomous driving system test.

[0038] In an exemplary embodiment, Figure 2 As shown, a method for generating an autonomous driving simulation scene is provided, which can be applied to electronic devices with electrical data processing capabilities, such as Figure 1 The server 102 shown, or the terminal computing device on the road side, the vehicle-mounted electronic device, etc., are not particularly limited in this embodiment of the present application. Figure 1 The server 102 in the example is used as an example to illustrate, including the following steps 202 to 208. Among them:

[0039] Step 202: Acquire multi-source data for autonomous driving simulation testing, perform path planning for the simulation scenario based on the multi-source data, and obtain a target planned path.

[0040] Among them, multi-source data includes: map data and intersection traffic flow data of each intersection. The intersection can be a crossroad with multiple traffic directions. The intersection traffic flow data at least includes: the real-time traffic conditions of the intersection, the behavior trajectory and density of traffic participants (including motor vehicles, non-motor vehicles and pedestrians). The map data at least includes: road geometry data, traffic element data (such as the location and status of traffic lights, traffic speed limit signs), data related to the road network, etc.

[0041] Specifically, the path planning of the simulation scenario can be: planning a long continuous path with expected mileage and continuity, aiming to solve the problem of how to plan the optimal driving path from the starting point to the end point of the autonomous driving vehicle in a complex road environment with multiple scenarios. The core challenge lies in the time length and scenario complexity of path planning, so that the target planning path not only covers different types of traffic scenarios, such as highways, urban roads, intersections, etc., but also covers a variety of traffic conditions (such as congestion, changes in traffic light signals, pedestrian and vehicle behavior, etc.).

[0042] Optionally, multi-source data for autonomous driving simulation testing can be obtained from the data background, and data preprocessing can be performed on the multi-source data. Data preprocessing includes but is not limited to: cleaning, format conversion and consistency checking of intersection traffic flow data and map data respectively to ensure the integrity, rationality and effectiveness of the data, providing a basis for subsequent analysis. Based on this, long-term continuous path planning can be performed based on the multi-source data after data preprocessing and in accordance with the preset path planning constraints to obtain a target planning path that covers as many and complete traffic scenarios as possible.

[0043] Step 204, determine at least one intersection in the target planning path, and screen the traffic flow segment data based on the intersection traffic flow data of at least one intersection in the multi-source data to determine the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene.

[0044] The passing intersection specifically refers to an intersection in the target planning path. The target planning path is composed of multiple passing intersections and road sections connecting the passing intersections. The intersection traffic flow data of each passing intersection may include: intersection traffic flow segment data corresponding to at least one traffic scene occurring at the passing intersection.

[0045] Alternatively, at least one intersection in the target planning path may be determined first, and the intersection traffic flow data of at least one intersection may be obtained from multi-source data, so that the intersection traffic flow data of at least one intersection may be screened according to the preset traffic flow segment data screening conditions, so as to screen out the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene from the intersection traffic flow data of at least one intersection. Specifically, meeting the conditions for generating the autonomous driving simulation scene may refer to: the intersection traffic flow segment data has a certain degree of traffic scene complexity and has traffic scene connectivity (the traffic scene is not isolated).

[0046] Step 206, for every two adjacent passing intersections in at least one passing intersection, based on the intersection traffic flow data of each adjacent passing intersection, traffic flow data of the connecting section with missing data between the adjacent passing intersections is inferred and restored to obtain the section traffic flow restoration data of the connecting section.

[0047] Among them, when the distance between adjacent passing intersections is far (such as greater than the preset distance), the collection equipment at the adjacent passing intersections cannot collect at least a part of the traffic flow data in the connecting section, resulting in data missing in the connecting section. In addition, when the traffic flow data of a passing intersection among the adjacent passing intersections is missing, it will also cause the intersection traffic flow data of the adjacent passing intersection to not completely cover the connecting section, resulting in data missing in the connecting section.

[0048] Alternatively, for every two adjacent passing intersections in at least one passing intersection, the connecting road section between the adjacent passing intersections can be detected based on the intersection traffic flow data of each adjacent passing intersection to determine whether there is data missing in the connecting road section between the adjacent passing intersections. That is, it is determined whether the intersection traffic flow data of the adjacent passing intersections has covered the connecting road section between the adjacent passing intersections. In the case of determining that there is data missing in the connecting road section between the adjacent passing intersections, the missing traffic flow data on the connecting road section can be further restored by traffic flow data reasoning based on the intersection traffic flow data of each adjacent passing intersection to ensure the spatiotemporal continuity of the traffic flow between the intersection and the road section, thereby obtaining the section traffic flow restoration data of the connecting road section.

[0049] In a possible implementation, take the example that intersection A and intersection B are adjacent and connected by section ab, the a end of section ab is connected to intersection A, the b end of section ab is connected to intersection B, and there is missing traffic flow data on section ab. The missing traffic flow data on connecting section ab can be inferred and restored based on the intersection traffic flow data of intersection A and intersection B, so that the traffic scenes occurring at intersection A, connecting section ab, and intersection B maintain spatiotemporal continuity.

[0050] Step 208, generating an autonomous driving simulation scene based on the intersection traffic flow data, the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions, and the road section traffic flow restoration data.

[0051] Among them, the autonomous driving simulation scenario covers multi-dimensional information such as road geometry, traffic signals, and participant behavior.

[0052] Exemplarily, the traffic flow data of at least one intersection passed through, the selected intersection traffic flow fragment data that meets the conditions for generating an autonomous driving simulation scene, and the road section traffic flow restoration data obtained by reasoning and restoration can be collected to form clear, continuous, complete and reasonable traffic flow data on the target planning path, so as to extract the autonomous driving simulation scene based on the collected traffic flow data and generate a continuous autonomous driving simulation scene.

[0053] Furthermore, to facilitate the use of autonomous driving simulation scenarios, key traffic scenes in the autonomous driving simulation scenarios can also be automatically labeled, including but not limited to: marking the behavior of traffic participants, road congestion, special traffic events, etc., to provide a reference for subsequent simulation analysis.

[0054] Furthermore, the annotated autonomous driving simulation scene can be stored in the format of a simulation scene file to facilitate the subsequent testing, verification and optimization of the autonomous driving system.

[0055] In a specific embodiment, the simulation scene file may include: (1) the starting coordinates of the vehicle in the current segment; ; (2) The starting state of the vehicle in the current segment: speed , acceleration , heading angle ; (3) The end coordinates of the vehicle in the current segment ; (4) The vehicle’s final state in the current segment: speed , acceleration , heading angle ; (5) Traffic participants in the current segment Driving trajectory , is the simulation time; (6) the traffic participants in the current segment Types of people, including but not limited to motor vehicles, non-motor vehicles, and pedestrians; (7) Traffic participants in the current segment The traffic participant object corresponding to the next segment ; (8) Traffic participants in the current segment Transition to the next segment.

[0056] The above-mentioned method for generating an autonomous driving simulation scene, after performing path planning according to multi-source data and obtaining a target planning path for the simulation scene, first determines at least one intersection in the target planning path, and then screens the traffic flow segment data according to the intersection traffic flow data of at least one intersection in the multi-source data, and determines the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions, so as to assist in generating the autonomous driving simulation scene. Further, for each two adjacent intersections in at least one intersection, the traffic flow data can be inferred and restored for the connecting road sections with missing data between the adjacent intersections according to the intersection traffic flow data of the adjacent intersections, so as to complete the traffic flow data on the target path and obtain the section traffic flow restoration data of the connecting road sections. Based on this, on the basis of the original intersection traffic flow data, combined with the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions and the section traffic flow restoration data obtained by inference and restoration, the traffic scene on the target planning path is made clearer, continuous and reasonable, which is conducive to generating an effective autonomous driving simulation scene based on the above traffic flow data, and thus improving the effect of autonomous driving simulation testing based on the autonomous driving simulation scene.

[0057] In one of the embodiments, path planning of a simulation scenario is performed based on multi-source data to obtain a target planned path, including: screening traversable areas based on map data in the multi-source data to obtain a topological map including traversable areas; obtaining path planning constraints of the simulation scenario, performing path planning according to the path planning constraints and the topological map to obtain at least one sub-planning path arranged in sequence; and connecting at least one sub-planning path in sequence to obtain a target planned path.

[0058] The drivable area specifically refers to: drivable road sections and drivable intersections. The path planning constraints can be flexibly configured according to the needs of the actual application scenario, so that the obtained target planning path can provide sufficient data for simulation testing.

[0059] Optionally, the passable area can be screened based on the map data in the multi-source data, closed roads or impassable areas can be excluded, and the drivable road sections and passable intersections can be screened out, so as to build a topological map based on the passable area. Further, the path planning constraints can be obtained to perform multiple rounds of long-term continuous path planning according to the path planning constraints and the topological map, and after each round of long-term continuous path planning, a path planning scheme can be obtained, and the path planning scheme can include at least one sub-planning path. In the process of multiple rounds of long-term continuous path planning, the next path planning scheme can be dynamically adjusted according to the last path planning scheme, and the end point of the last path planning scheme is made the starting point of the next path planning scheme. After collecting the path planning schemes obtained in all rounds, at least one sub-planning path arranged in sequence can be obtained. Based on this, at least one sub-planning path can be connected in sequence to obtain a continuous, long-term, target planning path covering multiple intersections.

[0060] For example, the topological map can be constructed in the following manner: the points before the entrance and after the exit in the traversable area are used as topological points. The construction position of the topological map points of the same intersection in the traversable area For the same macro node group, and based on the topological graph point Build connecting edges , connecting edges Can be characterized: from a certain intersection Or before an entrance , connected to an exit road .

[0061] In a specific example, Figure 3 As shown in FIG. 1 , a schematic diagram of connecting two rounds of path planning schemes to obtain a target planning path is provided. Among them, the target planning path includes passing intersections 1 to 6. The a1 end of the round path planning scheme is the entrance of the target planning path. The a2 end of the round path planning solution is The starting point of the round path planning scheme, The b2 end of the round path planning scheme is the exit of the target planning path.

[0062] In this embodiment, path planning can be used to obtain a target planning path formed by connecting at least one sub-planning path, so that the target planning path covers as rich and complete traffic scenes as possible, which is conducive to the subsequent screening, reasoning and restoration of traffic flow data on the target planning path, making the traffic scene on the target planning path clearer, continuous and reasonable, thereby generating an effective autonomous driving simulation scene and improving the effect of autonomous driving simulation testing based on the autonomous driving simulation scene.

[0063] In a specific example, the path planning constraints of the simulation scenario are obtained, including: constructing a path coverage evaluation function, and maximizing the value of the path coverage evaluation function as the first constraint; the value of the path coverage evaluation function represents the degree of diversity of sub-planning paths contained in a path; sequentially adjacent sub-planning paths satisfying the end-to-end connection relationship is used as the second constraint, and the total mileage of at least one sub-planning path reaches the expected mileage as the third constraint; the first constraint, the second constraint and the third constraint are combined to obtain the path planning constraints.

[0064] Among them, the setting of the path coverage evaluation function is to select a reasonable path to maximize the road coverage. The higher the coverage of the roads in the map by the sub-planning paths contained in a path, the higher the diversity of the sub-planning paths contained in a path, and the larger the value of the path coverage evaluation function. The sequentially adjacent sub-planning paths satisfy the head-to-tail connection relationship, which specifically means that the end point of the sub-planning path obtained by the previous round of path planning is the starting point of the sub-planning path obtained by the next round of path planning. The expected mileage can be configured based on the simulation mileage requirements and the planned simulation scenario type. It can be a small mileage such as 1km, 2km, or a large mileage such as 50km, 100km. The total mileage can reach the expected mileage and can be greater than or equal to the expected mileage. When the expected mileage is reached, the next round of path planning can be stopped and the target planning path can be output.

[0065] Specifically, in the process of performing multiple rounds of path planning, each path planning needs to satisfy the first constraint condition and the second constraint condition, and the path planning results of multiple rounds need to satisfy the third constraint condition.

[0066] In a specific embodiment, the goal of each round of long-term continuous path planning is to select a reasonable path planning solution. , It can represent the first Strip planning path.

[0067] First Taking the path planning as an example, the path coverage evaluation function involved in the first constraint condition can be specifically shown as function (1):

[0068] (1)

[0069] In function (1), is the round of path planning, Representation: Calculate the first to the The sum of the path coverage of the sub-planning paths contained in a path obtained by performing path planning for the first time, Characterization: Based on The path introduced by the path planning for the Tth time maximizes the path coverage of the path obtained after the path planning for the Tth time.

[0070] To calculate the A sub-planning path obtained by performing path planning Taking the path coverage of as an example, the coverage calculation method is shown in formula (2):

[0071] (2)

[0072] in, Characterization: This sub-planning path can be used as a connecting edge, from a certain intersection Or before an entrance , connected to an exit road . and is the coefficient configured according to actual needs. Representation Computing and The Euclidean distance between these two topological points, Characterize the calculation of the last round of path planning The number of occurrences. random is a random number with a (0, 1] distribution, which is used to reduce repeated paths and introduce random items to increase diversity.

[0073] Furthermore, the second constraint condition can be specifically expressed as formula (3):

[0074] (3)

[0075] in, Characterization The starting point of the path planning round, Characterization The end point of the path planning round. The second constraint is used to constrain: the end point of the previous path planning solution is the starting point of the next path planning solution. Based on this, multiple rounds of path planning can be used to obtain multiple sub-planning paths arranged in sequence, and multiple sub-planning paths are connected to obtain a continuous target planning path.

[0076] In this embodiment, long-term continuous path planning can be reasonably performed through multiple constraints, so as to obtain a target planning path that covers as rich and complete traffic scenarios as possible.

[0077] In some embodiments, traffic flow segment data is screened based on the intersection traffic flow data of at least one passing intersection in multi-source data to determine the intersection traffic flow segment data that meets the conditions for generating an autonomous driving simulation scene, including: obtaining the intersection traffic flow data of at least one passing intersection from the multi-source data, the intersection traffic flow data including: intersection traffic flow segment data corresponding to at least one traffic scene; for the intersection traffic flow data of each passing intersection, performing traffic scene complexity analysis and traffic scene connectivity analysis on at least one intersection traffic flow segment data in the intersection traffic flow data, respectively, to obtain a scene evaluation result of at least one intersection traffic flow segment data; based on the scene evaluation result of at least one intersection traffic flow segment data, determining the intersection traffic flow segment data that meets the conditions for generating an autonomous driving simulation scene from at least one intersection traffic flow segment data.

[0078] Alternatively, the intersection traffic flow data of at least one of the passing intersections can be obtained from the multi-source data. The intersection traffic flow data includes: intersection traffic flow segment data corresponding to at least one traffic scene occurring at the passing intersection, which can specifically include: the behavior, trajectory, density, etc. of the traffic participants, and the intersection traffic flow data carries a timestamp.

[0079] Furthermore, for each intersection traffic flow data passing through the intersection, the traffic scene complexity analysis and traffic scene connectivity analysis can be performed on at least one intersection traffic flow segment data in the intersection traffic flow data, so as to obtain the scene evaluation results of at least one intersection traffic flow segment data. Taking the traffic scene complexity analysis as an example, it includes but is not limited to: analyzing the density and type of traffic participants, the state of traffic elements (including but not limited to traffic lights), and the traffic time period in the intersection traffic flow segment data within a certain period of time, so as to evaluate the traffic scene complexity of the traffic scene represented by the intersection traffic flow segment data. Taking the traffic scene connectivity analysis as an example (the intersection traffic flow segment data can determine the order according to the timestamps carried), it includes but is not limited to: (1) analyzing whether the starting point of the traffic scene represented by the current intersection traffic flow segment data is on the same road as the end point of the traffic scene represented by the previous intersection traffic flow segment data; (2) analyzing whether the end point of the traffic scene represented by the current intersection traffic flow segment data is on the same road as the starting point of the traffic scene represented by the next intersection traffic flow segment data, so as to evaluate the traffic scene connectivity of the traffic scene represented by the intersection traffic flow segment data.

[0080] Furthermore, for each intersection traffic flow data passing through the intersection, after obtaining the scene evaluation results of at least one intersection traffic flow segment data in the intersection traffic flow data, the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene can be determined from the at least one intersection traffic flow segment data based on the scene evaluation results of the at least one intersection traffic flow segment data. That is, the intersection traffic flow segment data whose traffic scene complexity reaches the preset complexity and whose traffic scene has connectivity is selected, so that the traffic scene represented by the selected intersection traffic flow segment data is sufficiently diverse and complex.

[0081] In a specific example, when the traffic scene represented by the intersection traffic flow segment data meets at least one of the following conditions, it can be determined that the traffic scene complexity of the traffic scene represented by the intersection traffic flow segment data reaches a preset complexity: the density of traffic participating objects in the intersection traffic flow segment data reaches a preset density within a certain period of time, the type of traffic participating objects in the intersection traffic flow segment data within a certain period of time is a preset type, the state of traffic elements in the intersection traffic flow segment data within a certain period of time is a preset state, and the current traffic time period is in a peak travel period.

[0082] In an optional embodiment, when the traffic scene represented by the intersection traffic flow segment data meets the following two conditions, it can be determined that the traffic scene represented by the intersection traffic flow segment data has connectivity: (1) the starting point of the traffic scene represented by the current intersection traffic flow segment data is on the same road as the end point of the traffic scene represented by the previous intersection traffic flow segment data; (2) the end point of the traffic scene represented by the current intersection traffic flow segment data is on the same road as the starting point of the traffic scene represented by the next intersection traffic flow segment data.

[0083] In this embodiment, traffic scene complexity analysis and traffic scene connectivity analysis can be performed on the intersection traffic flow data passing through the target planning path, and intersection traffic flow fragment data that covers both complex traffic scenes and has traffic scene connectivity can be screened out. The screened intersection traffic flow fragment data can then be combined to generate an autonomous driving simulation scene covering rich and complex traffic scenes, thereby improving the autonomous driving simulation test effect.

[0084] In an exemplary embodiment, based on the intersection traffic flow data of each adjacent passing intersection, the traffic flow data of the connecting section with missing data between the adjacent passing intersections is inferred and restored to obtain the section traffic flow restoration data of the connecting section, including: based on the intersection traffic flow data of each adjacent passing intersection, the data flow integrity test is performed on the connecting section between the adjacent passing intersections to obtain the test result; when the test result indicates that there is missing data in the connecting section, the movement trajectory of the traffic participants in the connecting section is inferred and restored based on the intersection traffic flow data of each adjacent passing intersection to obtain the section traffic flow restoration data of the connecting section.

[0085] Optionally, for each connecting section between adjacent passing intersections, data flow integrity detection can be performed on the connecting section between adjacent passing intersections based on the intersection traffic flow data of each adjacent passing intersection to detect whether there is data missing in the connecting section between adjacent passing intersections and obtain a detection result. In the case where the detection result indicates that there is data missing in the connecting section, the motion trajectory of the traffic participants in the connecting section can be reasonably inferred and restored based on the intersection traffic flow data of each adjacent passing intersection to restore the missing traffic flow data and obtain the section traffic flow restoration data of the connecting section, thereby ensuring the spatiotemporal continuity of the traffic flow between the connecting section and the adjacent passing intersection to which it belongs.

[0086] In a possible implementation, a two-stage reasoning and restoration scheme may be adopted in this embodiment. Specifically, in the first stage, for adjacent passing intersections, the traffic participants whose motion trajectories appear in the connecting section may be determined based on the intersection traffic flow data of each adjacent passing intersection, that is, the traffic participants whose motion trajectories can be reasoned and restored are first determined. In the second stage, the motion trajectories of the determined traffic participants are reasoned and restored in combination with the intersection traffic flow data of each adjacent passing intersection.

[0087] In a specific example, Figure 4 As shown, a schematic diagram of the motion trajectory obtained by inferring and restoring traffic flow data is provided. Among them, the traffic scene represented by the intersection traffic flow fragment data at the end time of the previous segment includes at least a part of the motion trajectory of vehicle A1, vehicle A2, and vehicle A3. The intersection traffic flow fragment data at the start time of the next segment includes at least a part of the motion trajectory of vehicle B1, vehicle B2, vehicle B3, and vehicle B4. Based on this, the missing traffic flow data can be inferred and restored based on the intersection traffic flow fragment data at the end time of the previous segment and the intersection traffic flow fragment data at the start time of the next segment, so as to Figure 4 As shown, the motion trajectory of vehicle A1 becoming vehicle B1 by driving, the motion trajectory of vehicle A2 becoming vehicle B2 by driving, and the motion trajectory of vehicle A3 becoming vehicle B3 by driving are restored to make the entire traffic scene continuous.

[0088] It should be noted that after reasonably reasoning and restoring the motion trajectories of traffic participants in the connecting sections and obtaining the missing section traffic flow restoration data, it is possible to further verify whether the section traffic flow restoration data is reasonable, that is, to check the connectivity of adjacent passing intersections. For example, verify whether the traffic scene represented by the section traffic flow restoration data makes the traffic scenes in adjacent passing intersections and connecting sections continuous and reasonable.

[0089] Take the example that intersection A and intersection B are adjacent and connected by section ab, the a end of section ab is connected to intersection A, and the b end of section ab is connected to intersection B. If data missing is detected at the a end and the middle of section ab, and the traffic flow restoration data of the missing sections at the a end and the middle of section ab are obtained through inference and restoration, it can be further verified that: (1) the traffic scene represented by the traffic flow restoration data at the a end is consistent with the traffic scene at intersection A; (2) the traffic scene represented by the traffic flow restoration data in the middle of section can make the overall traffic scene between intersection A, section ab, and intersection B continuous, consistent, and reasonable. If anomalies are verified, the inference and restoration can be repeated, or feedback can be sent to manual processing to obtain the traffic flow restoration data of the section that keeps the overall traffic scene free of anomalies.

[0090] In this embodiment, for connecting sections where data is missing between adjacent intersections, the motion trajectories of traffic participants in the connecting sections can be inferred and restored to complete the traffic flow data on the target path, making the traffic scene on the target planned path more continuous and reasonable, which is conducive to the subsequent combination of the section traffic flow restoration data obtained after reasoning and restoration to generate an autonomous driving simulation scene covering rich and complex traffic scenes, thereby improving the effect of autonomous driving simulation testing.

[0091] In a possible implementation, the motion trajectories of traffic participants in the connecting section are inferred and restored based on the intersection traffic flow data of each adjacent passing intersection to obtain the section traffic flow restoration data of the connecting section, including: determining the first intersection and the second intersection in the adjacent passing intersections, as well as the first traffic participant object that disappears at the first intersection and the second traffic participant object that appears at the second intersection based on the intersection traffic flow data of each adjacent passing intersection; the first intersection is connected to the second intersection via the connecting section; the first traffic participant object and the second traffic participant object are paired to obtain at least one matching pair; based on the intersection traffic flow data of each of the first intersection and the second intersection, the motion trajectories of the traffic participants in at least one matching pair are inferred and restored to obtain the section traffic flow restoration data of the connecting section.

[0092] Optionally, in the case where there is missing data (partial or complete missing) in the connecting sections of adjacent intersections, a two-stage reasoning and restoration scheme can be used to reconstruct the traffic flow data of traffic participants in a continuous traffic scenario across intersections, and obtain the traffic flow restoration data of the missing sections.

[0093] Specifically, in the first stage, the first intersection (which can be regarded as the previous intersection) and the second intersection (which can be regarded as the next intersection) in the adjacent passing intersections can be determined based on the intersection traffic flow data of each adjacent passing intersection, as well as the first traffic participant object that disappears at the first intersection and the second traffic participant object that appears at the second intersection. Further, a matching pair of the first traffic participant object and the second traffic participant object, and a matching pair set including the constructed matching pairs can be preliminarily constructed. Further, a first multi-objective optimization function for optimizing the matching pair set can be constructed to optimize the matching relationship between the first traffic participant object and the second traffic participant object in the matching pair set, and finally achieve the pairing of the first traffic participant object and the second traffic participant object to obtain at least one matching pair. In the second stage, based on the intersection traffic flow data of each of the first intersection and the second intersection, the motion trajectory of the traffic participant object in at least one matching pair can be inferred and restored respectively through the second multi-objective optimization function for optimizing the motion trajectory of the traffic participant object, so as to obtain the section traffic flow restoration data of the connecting section.

[0094] In a possible implementation, taking the first stage as an example, the first traffic participant object can be preliminarily constructed and the second traffic participant Matching pairs between , and by matching pairs The set of matching pairs , and construct a set of optimized matching pairs The first multi-objective optimization function is to maximize the sum of the interpolation trajectory point spacings of all matching pairs of traffic participants at the same time, and minimize the sum of the interpolation trajectory speed changes of all matching pairs of traffic participants. Based on this, the first multi-objective optimization function can be specifically shown as function (4):

[0095] (4)

[0096] in, Can be characterized: Matching pairs at the moment The interpolation trajectory point spacing of traffic participants, Can characterize: matching pairs The speed change of the interpolated trajectory of the traffic participant in the experiment includes but is not limited to: the absolute value of the difference between the average speed and the starting speed and the ending speed.

[0097] Furthermore, in the second stage, according to the intersection traffic flow data of the first intersection and the second intersection, based on the matching pairs obtained by the optimization in the first stage, a second multi-objective optimization function for optimizing the motion trajectory of traffic participants can be constructed, so as to infer and restore the motion trajectory of the traffic participants in at least one matching pair, generate a trajectory set S, and obtain the traffic flow restoration data of the connecting section. The trajectory set S (including the trajectories s of the traffic participants in multiple matching pairs) satisfies: (1) at the same time, the sum of the trajectory point spacing of the traffic participants in all matching pairs is maximized; (2) at continuous moments, the change in trajectory speed of the traffic participants in all matching pairs is minimized; (3) at the same time, the lane line offset of the trajectories of the traffic participants in all matching pairs is minimized. Based on this, the second multi-objective optimization function can be specifically shown as function (5):

[0098] (5)

[0099] In function (5), Characterization: Computation At this moment, matching The distance between the trajectory points of the traffic participants under trajectory s; Representation: Counting consecutive moments Next, match the The change in trajectory speed of the traffic participant under trajectory s is to constrain the speed of the traffic participant to meet the starting point constraint and the end point constraint and not to change suddenly; Characterization: Computation At this moment, matching Traffic participants in the trajectory Left lane line below and right lane line The distance difference, that is, the trajectory lane line offset is calculated to ensure that the vehicle travels normally on the road.

[0100] It should be noted that, in the process of performing multi-objective optimization based on the second multi-objective optimization function (5), kinematic constraints must be met. The kinematic constraints may specifically include:

[0101] , to constrain lateral motion, is the time interval;

[0102] , to constrain longitudinal motion;

[0103] , to constrain the plane rotation, ;

[0104] It can be configured according to actual needs to ensure fast tracking of the target vertically;

[0105] Can be configured according to actual needs to limit the front wheel turning angle;

[0106] It can be configured according to actual needs to avoid excessive steering.

[0107] In this embodiment, for each adjacent passing intersection, the motion trajectory of the traffic participant objects that are successfully paired in the adjacent passing intersections can be inferred and restored, so that the motion trajectory of the traffic participant objects between the adjacent passing intersections is more reasonable and continuous, thereby making the traffic scenes on the connecting sections between the adjacent passing intersections richer, more reasonable and continuous, which is conducive to the subsequent combination of the road section traffic flow restoration data obtained after reasoning and restoration, to generate an autonomous driving simulation scene covering rich and complex traffic scenes, thereby improving the effect of autonomous driving simulation testing.

[0108] In a specific example, based on the above embodiments, Figure 5 As shown, a flowchart of another method for generating a simulation scene for autonomous driving is provided, which mainly includes the following steps:

[0109] (1) Multi-source data import and preprocessing, including: importing multi-source data (multi-source heterogeneous data) from the data background, and preprocessing the multi-source data, such as cleaning, format conversion and consistency check, and performing data rationality analysis through consistency check. The data background can store the intersection traffic flow data of each intersection obtained from the cloud control platform and the map data obtained from the map system for import.

[0110] (2) Long-term continuous path planning, including: selecting a topological map from the map data based on the map data in the multi-source data and the configured expected mileage, and performing long-term continuous path planning for the simulation scenario based on the topological map and the preset path planning constraints to obtain the target planned path.

[0111] (3) Intersection traffic flow segment data screening, including: determining at least one intersection in the target planning path, and screening the traffic flow segment data based on the intersection traffic flow data of at least one intersection in the multi-source data, to determine the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scenario.

[0112] (4) Reasoning and restoring section traffic flow data, including: for each two adjacent intersections in at least one intersection, performing a two-stage multi-objective optimization based on the intersection traffic flow data of the adjacent intersections and a pre-constructed first multi-objective optimization function and a pre-constructed second multi-objective optimization function, so as to reason and restore the traffic flow data of the connecting section where there are missing data between the adjacent intersections, and obtain the section traffic flow restoration data of the connecting section.

[0113] (5) Generation of continuous autonomous driving simulation scene data, including: generating autonomous driving simulation scenes based on intersection traffic flow data, intersection traffic flow segment data that meets the conditions for generating autonomous driving simulation scenes, and road section traffic flow restoration data.

[0114] (6) Convert the autonomous driving simulation scenario into a format acceptable to the autonomous driving simulation platform and store it in the form of a simulation file in a background database connected to the server as a long-term data asset for subsequent simulation testing, scenario reproduction, or scenario mining and improvement, thereby forming a series of simulation scenario libraries that can be called by corporate customers.

[0115] In a specific application scenario, an algorithm tool chain for an autonomous driving simulation scenario library can be constructed based on the above-mentioned autonomous driving simulation scenario generation method. Specifically, based on the real traffic flow data obtained on the roadside of the demonstration area, the generated continuous autonomous driving simulation scenario can be verified through the autonomous driving simulation platform to test the performance of the autonomous driving algorithm in different scenarios, such as focusing on analyzing collision risks, path planning effects, vehicle control, etc., to provide basic scenario data and an algorithm tool chain for scenario analysis for the construction of the simulation scenario library system of the demonstration area.

[0116] Based on the above-mentioned method for generating autonomous driving simulation scenarios, the data collected weekly from 12 intersections in a certain intelligent network demonstration area is used as an example to generate autonomous driving simulation scenarios. After obtaining the autonomous driving simulation scenarios, the autonomous driving algorithm can be injected into the autonomous driving simulation scenarios to train the ability of the autonomous driving algorithm to find problems, which can effectively reproduce the failure phenomenon of the autonomous driving algorithm. The specific failure phenomenon reproduced can be shown in Table 1:

[0117] Table 1 Display of the reproducible failure phenomena

[0118]

[0119] The following test conclusions can be obtained: (1) In terms of generation capability, the maximum mileage of the automatically generated continuous autonomous driving simulation scene is ≥100 km; (2) In terms of scene coverage, the road coverage of the automatically generated continuous autonomous driving simulation scene is ≥90%; (3) In terms of usage effect, the recurrence rate of autonomous driving simulation problems is ≥80%.

[0120] In this embodiment, based on the above-mentioned autonomous driving simulation scenario generation method, an algorithm tool chain for the autonomous driving simulation scenario library is constructed, which can effectively help verify and test the autonomous driving algorithm and improve the stability and overall safety of the autonomous driving system.

[0121] The method for generating autonomous driving simulation scenarios based on the above embodiments can be applied to the following application scenarios / products: (1) Application to traffic scenario library products: This method can be used to build a continuous traffic scenario library covering long-term driving, and store the generated continuous scenarios containing multiple intersections and sections in a database. Automakers or simulation platforms can use these scenario libraries to test autonomous driving systems, especially to evaluate the global performance of autonomous driving systems in different roads, traffic flows and complex environments, so as to improve the stability and overall safety of autonomous driving systems. (2) Application to autonomous driving data processing and analysis tools: This method is used to convert real traffic data into a long-term continuous scenario format that can be used for simulation, and perform scenario reasoning and completion, so that it can be applied to the vehicle development process, help the autonomous driving system to handle traffic scenarios in the real world, and optimize the vehicle's global path planning, traffic flow prediction and decision-making capabilities. (3) Application to autonomous driving algorithm verification tools: Autonomous driving companies can design the processing flow of this method into a continuous scenario generation tool for testing the perception, decision-making and control algorithms of autonomous driving systems in long-term driving. By generating continuous scenarios containing multiple sections and intersections, the performance of the autonomous driving system in different situations can be more comprehensively verified to ensure its stability and adaptability in the actual road environment. (4) Application to autonomous driving simulation platform: This method can be integrated into the autonomous driving simulation platform to generate long-term, continuous simulation scenarios to help verify and test the global performance of the autonomous driving system. By generating real scenarios covering multiple intersections and road sections, the autonomous driving system can be operated and tested in various complex continuous traffic flow scenarios, such as long-distance driving, urban roads, and multiple intersections, ensuring its safety and reliability in global driving.

[0122] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0123] Based on the same inventive concept, the embodiment of the present application also provides an autonomous driving simulation scene generation device for implementing the autonomous driving simulation scene generation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more autonomous driving simulation scene generation device embodiments provided below can refer to the limitations of the autonomous driving simulation scene generation method above, and will not be repeated here.

[0124] In an exemplary embodiment, Figure 6 As shown, an automatic driving simulation scene generation device is provided, comprising: a target path planning module 602, a road intersection traffic flow data screening module 604, a road section traffic flow data restoration module 606 and a simulation scene generation module 608, wherein:

[0125] The target path planning module is used to obtain multi-source data for autonomous driving simulation testing, perform path planning for simulation scenarios based on the multi-source data, and obtain the target planning path. The multi-source data includes traffic flow data and map data.

[0126] The intersection traffic flow data screening module is used to determine at least one intersection in the target planning path, and screen the traffic flow segment data according to the intersection traffic flow data of at least one intersection in the multi-source data, and determine the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene;

[0127] A road section traffic flow data restoration module is used to perform traffic flow data inference restoration on a connecting road section where data is missing between two adjacent passing intersections in at least one passing intersection according to the intersection traffic flow data of the adjacent passing intersections, so as to obtain the road section traffic flow restoration data of the connecting road section;

[0128] The simulation scenario generation module is used to generate an autonomous driving simulation scenario based on intersection traffic flow data, intersection traffic flow segment data that meets the autonomous driving simulation scenario generation conditions, and road section traffic flow restoration data.

[0129] The above-mentioned automatic driving simulation scene generation device, after performing path planning according to multi-source data and obtaining the target planning path of the simulation scene, first determines at least one passing intersection in the target planning path, and then screens the traffic flow segment data according to the intersection traffic flow data of at least one passing intersection in the multi-source data, and determines the intersection traffic flow segment data that meets the automatic driving simulation scene generation conditions, so as to assist in the generation of the automatic driving simulation scene. Further, for each two adjacent passing intersections in at least one passing intersection, the traffic flow data can be inferred and restored for the connecting road sections with missing data between the adjacent passing intersections according to the intersection traffic flow data of the adjacent passing intersections, so as to complete the traffic flow data on the target path and obtain the section traffic flow restoration data of the connecting road sections. Based on this, on the basis of the original intersection traffic flow data, combined with the intersection traffic flow segment data that meets the automatic driving simulation scene generation conditions and the section traffic flow restoration data obtained by inference and restoration, the traffic scene on the target planning path is made clearer, continuous and reasonable, which is conducive to generating an effective automatic driving simulation scene based on the above traffic flow data, and thus improving the effect of automatic driving simulation testing based on the automatic driving simulation scene.

[0130] In one of the embodiments, the target path planning module is also used to: screen the traversable area based on the map data in the multi-source data to obtain a topological map including the traversable area; obtain the path planning constraints of the simulation scenario, perform path planning according to the path planning constraints and the topological map to obtain at least one sub-planning path arranged in sequence; connect at least one sub-planning path in sequence to obtain the target planning path.

[0131] In a specific example, the autonomous driving simulation scene generation device also includes a path planning constraint condition acquisition module, which is also used to: construct a path coverage evaluation function, and maximize the value of the path coverage evaluation function as a first constraint condition; the value of the path coverage evaluation function represents the degree of diversity of sub-planning paths contained in a path; the sequentially adjacent sub-planning paths satisfying the end-to-end connection relationship is used as a second constraint condition, and the total mileage of at least one sub-planning path reaches the expected mileage as a third constraint condition; the first constraint condition, the second constraint condition and the third constraint condition are combined to obtain the path planning constraint condition.

[0132] In some embodiments, the intersection traffic flow data screening module is also used to: obtain intersection traffic flow data of at least one intersection passing through from multi-source data, the intersection traffic flow data including: intersection traffic flow segment data corresponding to at least one traffic scene; for each intersection traffic flow data passing through the intersection, perform traffic scene complexity analysis and traffic scene connectivity analysis on at least one intersection traffic flow segment data in the intersection traffic flow data, and obtain a scene evaluation result of at least one intersection traffic flow segment data; based on the scene evaluation result of at least one intersection traffic flow segment data, determine the intersection traffic flow segment data that meets the conditions for generating an autonomous driving simulation scene from at least one intersection traffic flow segment data.

[0133] In an exemplary embodiment, the road section traffic flow data restoration module is also used to: perform data flow integrity detection on the connecting sections between adjacent passing intersections based on the intersection traffic flow data of each adjacent passing intersection to obtain a detection result; when the detection result indicates that there is data missing in the connecting section, the motion trajectory of the traffic participants in the connecting section is inferred and restored based on the intersection traffic flow data of each adjacent passing intersection to obtain the section traffic flow restoration data of the connecting section.

[0134] In a possible implementation, the section traffic flow data restoration module can also be used to: determine the first intersection and the second intersection in adjacent passing intersections, as well as the first traffic participant object that disappears at the first intersection and the second traffic participant object that appears at the second intersection based on the intersection traffic flow data of each adjacent passing intersection; the first intersection is connected to the second intersection via a connecting section; the first traffic participant object and the second traffic participant object are paired to obtain at least one matching pair; based on the intersection traffic flow data of each of the first intersection and the second intersection, the motion trajectories of the traffic participant objects in at least one matching pair are inferred and restored to obtain the section traffic flow restoration data of the connecting section.

[0135] Each module in the above-mentioned automatic driving simulation scene generation device can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0136] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data for generating automatic driving simulation scenarios. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for generating an automatic driving simulation scenario is implemented.

[0137] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0138] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0139] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0140] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0141] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0142] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0143] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0144] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for generating an autonomous driving simulation scene, characterized in that: The method comprises: Acquire multi-source data for autonomous driving simulation testing, perform path planning for a simulation scenario based on the multi-source data, and obtain a target planned path, wherein the multi-source data includes traffic flow data and map data; Determine at least one intersection in the target planning path, and screen traffic flow segment data according to the intersection traffic flow data of the at least one intersection in the multi-source data to determine the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene; For each two adjacent passing intersections among the at least one passing intersection, based on the intersection traffic flow data of each of the adjacent passing intersections, traffic flow data inference restoration is performed on a connecting road section with missing data between the adjacent passing intersections to obtain section traffic flow restoration data of the connecting road section; An autonomous driving simulation scene is generated based on the intersection traffic flow data, the intersection traffic flow segment data that meets the conditions for generating an autonomous driving simulation scene, and the road section traffic flow restoration data.

2. The method according to claim 1, characterized in that The path planning of the simulation scene based on the multi-source data to obtain a target planning path includes: Screening the traversable area based on the map data in the multi-source data to obtain a topological map including the traversable area; Obtaining path planning constraints of the simulation scenario, performing path planning according to the path planning constraints and the topological map, and obtaining at least one sub-planning path arranged in sequence; The at least one sub-planning path is connected in sequence to obtain a target planning path.

3. The method according to claim 2, characterized in that The obtaining of the path planning constraint conditions of the simulation scenario includes: Constructing a path coverage evaluation function, and taking the maximization of the value of the path coverage evaluation function as the first constraint condition; the value of the path coverage evaluation function represents the degree of diversity of the sub-planning paths contained in a path; The second constraint condition is that the sequentially adjacent sub-planning paths satisfy the head-to-tail connection relationship, and the third constraint condition is that the total mileage of at least one sub-planning path reaches the expected mileage; The first constraint condition, the second constraint condition and the third constraint condition are collected to obtain the path planning constraint condition.

4. The method according to claim 1, characterized in that: The filtering of traffic flow segment data according to the intersection traffic flow data of at least one of the intersections passed through in the multi-source data to determine the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene includes: Obtaining the intersection traffic flow data of each of the at least one passing intersection from the multi-source data; the intersection traffic flow data includes: intersection traffic flow segment data corresponding to at least one traffic scene; For each of the intersection traffic flow data, a traffic scene complexity analysis and a traffic scene connectivity analysis are respectively performed on at least one intersection traffic flow segment data in the intersection traffic flow data to obtain a scene evaluation result of each of the at least one intersection traffic flow segment data; Based on the respective scenario evaluation results of the at least one intersection traffic flow segment data, determine the intersection traffic flow segment data that meets the conditions for generating an autonomous driving simulation scenario from the at least one intersection traffic flow segment data.

5. The method according to claim 1, characterized in that The method of performing traffic flow data inference and restoration on a connecting road section with missing data between the adjacent passing intersections according to the respective intersection traffic flow data of the adjacent passing intersections to obtain the section traffic flow restoration data of the connecting road section comprises: According to the intersection traffic flow data of each of the adjacent passing intersections, a data flow integrity test is performed on the connecting section between the adjacent passing intersections to obtain a test result; When the detection result indicates that there is missing data in the connecting section, the motion trajectories of the traffic participants in the connecting section are inferred and restored based on the intersection traffic flow data of each of the adjacent passing intersections to obtain the section traffic flow restoration data of the connecting section.

6. The method according to claim 5, characterized in that The method of inferring and restoring the motion trajectory of the traffic participants in the connecting section according to the intersection traffic flow data of each adjacent passing intersection to obtain the section traffic flow restoration data of the connecting section includes: Determine, according to the intersection traffic flow data of each of the adjacent passing intersections, a first intersection and a second intersection among the adjacent passing intersections, as well as a first traffic participant object that disappears at the first intersection and a second traffic participant object that appears at the second intersection; the first intersection is connected to the second intersection via the connecting section; Pairing the first traffic participant object and the second traffic participant object to obtain at least one matching pair; According to the intersection traffic flow data of the first intersection and the second intersection respectively, the motion trajectories of the traffic participants in the at least one matching pair are inferred and restored respectively to obtain the section traffic flow restoration data of the connecting section.

7. An automatic driving simulation scene generation device, characterized in that: The device comprises: A target path planning module is used to obtain multi-source data for autonomous driving simulation testing, perform path planning for simulation scenarios based on the multi-source data, and obtain a target planned path, wherein the multi-source data includes traffic flow data and map data; An intersection traffic flow data screening module is used to determine at least one intersection in the target planning path, and screen traffic flow segment data according to the intersection traffic flow data of the at least one intersection in the multi-source data, to determine the intersection traffic flow segment data that meets the conditions for generating the autonomous driving simulation scene; A road section traffic flow data restoration module is used to perform traffic flow data inference restoration on a connecting road section where data is missing between the adjacent passing intersections according to the intersection traffic flow data of each of the adjacent passing intersections for each two adjacent passing intersections in the at least one passing intersection, so as to obtain the road section traffic flow restoration data of the connecting road section; The simulation scene generation module is used to generate an autonomous driving simulation scene based on the intersection traffic flow data, the intersection traffic flow segment data that meets the autonomous driving simulation scene generation conditions, and the road section traffic flow restoration data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images