A method and system for automatically generating a static map in a simulation test environment and a storage medium
By clustering and Markov property analysis of real road data and combining Monte Carlo methods to generate simulation maps, the problem of unrealistic simulation map generation in existing technologies is solved, and large-scale, efficient simulation map generation and autonomous driving algorithm testing are realized.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHONGQING CHANGAN AUTOMOBILE CO LTD
- Filing Date
- 2023-05-25
- Publication Date
- 2026-07-24
AI Technical Summary
Existing simulation map generation methods cannot generate maps on a large scale and the road topology is not realistic, resulting in poor reliability for autonomous driving simulation testing.
By dividing and clustering real-world road data, road features are constructed using Gaussian mixture models and Markov properties. The Monte Carlo method is then used to generate simulated road maps, ensuring complete road features and realistic topology.
It enables large-scale generation of simulated map topologies with real road features, improving the efficiency and coverage of iterative testing for autonomous driving algorithms.
Smart Images
Figure CN116805086B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of autonomous driving technology, and more specifically, to a method, system, and storage medium for automatically generating static maps in a simulation testing environment. Background Technology
[0002] Currently, existing methods for generating simulated maps generally rely on pre-defined map modules and the road features represented by each module. Multiple simulated maps are constructed by combining some or all of these modules with the road features. Because these maps are built using limited map modules and incomplete road features, they are not suitable for large-scale generation. Furthermore, the generated maps differ significantly from real-world road topology, and these unrealistic roads hinder the reliance of autonomous driving simulation testing on real roads. Therefore, it is crucial to develop automated simulated road generation technology that generates large-scale simulated maps with complete road features and realistic road topology to accelerate the evaluation of autonomous driving algorithms. Summary of the Invention
[0003] To address the problems of existing technologies that cannot generate simulation maps on a large scale and produce unrealistic road topologies, this invention provides a method, system, and storage medium for automatically generating static maps in a simulation testing environment. This method can construct simulation maps with complete road features and realistic road topologies, and can automatically generate simulation maps on a large scale.
[0004] To achieve the above-mentioned objectives of this invention, the technical solution adopted is as follows:
[0005] A method for automatically generating static maps in a simulation test environment, the method comprising the following steps:
[0006] The real basic road data is divided into several first-class roads, and the feature sets of the first-class roads corresponding to the first-class roads are extracted;
[0007] A Gaussian mixture model is used to cluster several first-class roads into several second-class roads, forming a second road feature set, wherein the second road feature set includes several first-class road feature sets.
[0008] The first road feature set in the second road feature set is resampled, and then the expected value of the first road feature set is calculated. The first road feature set with an expected value greater than a set threshold is selected to generate the road starting point.
[0009] By statistically analyzing the Markov properties of road features in real basic road data, the transition probabilities between various road features are constructed, and a simulated map of roads is generated by combining the road starting points.
[0010] The Monte Carlo method is used to evaluate the existence probability f of the roads in the generated simulation map, and road starting points are generated on a large scale to achieve the final generation of roads in the simulation map.
[0011] Preferably, the road features include road curvature, lateral slope, longitudinal slope, number of lanes, whether it is an intersection, intersection shape, lane line shape, lane line color, tunnels, and bridges.
[0012] Preferably, the calculation formula for the Gaussian mixture model is as follows:
[0013]
[0014] Where Θ=[K,ω k ,θ k K represents the number of Gaussian kernels in the Gaussian mixture model, which is also the number of first-class roads; ω k This represents the weight of the k-th Gaussian component in the Gaussian mixture model, and the sum of all weights equals 1. θ k =[μ k ,∑ k ] represents the mean matrix corresponding to the k-th Gaussian component ∑ k sum of variance matrix μ k ;x GMM Represents the road features of the input:
[0015] x GMM =[ρ,θ lat ,θ lon ,η,I,R,S,C,T,B] T (2)
[0016] Where ρ represents the road curvature, θ lat θ represents the lateral slope. lon The values represent the longitudinal slope, η represents the number of lanes, I represents whether it is an intersection, R represents the intersection shape, S represents the lane line shape, C represents the lane line color, T represents a tunnel, and B represents a bridge.
[0017] Furthermore, the number of Gaussian kernels in the Gaussian mixture model is calculated as follows;
[0018] K min =arg min K (Kln(n c )-2ln(log(f(x GMM ;Θ)))) (3)
[0019] Where, arg min K This indicates finding the minimum value, where K represents the number of Gaussian kernels in the Gaussian mixture model, and n... cIndicates the amount of data used for clustering, x GMM K represents the input road features. min This represents the number of minimum Gaussian kernels obtained by solving the problem.
[0020] Preferably, the formula for calculating the starting point of the road is as follows:
[0021]
[0022] Where, x i =[ρ,θ lat ,θ lon ,η,I,R,S,C,T,B] T Let ρ represent the road feature parameters sampled in the i-th time, and θ represent the road curvature. lat θ represents the lateral slope. lon The values represent the longitudinal slope, η represents the number of lanes, I represents whether it is an intersection, R represents the intersection shape, S represents the lane line shape, C represents the lane line color, T represents a tunnel, and B represents a bridge. x represents the sampled value i The value is given by n, which represents the number of samples, and E represents the expected value.
[0023] when When the value is greater than the set threshold, select the corresponding x. i Generate the starting point of the road.
[0024] Preferably, the transition probabilities between each road feature are constructed as follows:
[0025]
[0026] Where, π α,β M represents the number of times road feature α is transferred to road feature β. α Represents road characteristic α;
[0027] Therefore, the state transition probability matrix π = [π] is obtained. α,β ] N*N , where N represents the size of the matrix.
[0028] Preferably, the transition probabilities between various road features are combined with the road starting point to construct a simulated road with real road characteristics, as shown in the following expression:
[0029] P(M α )=P(π α,β ,x i ), i = 1, 2, ..., n (6)
[0030] Among them, M α This refers to the roads in the generated simulation map; P(M) α ) represents road M in the simulation map.α The road feature sequence, x i Let P represent the starting point of the road and P represent the probability.
[0031] The expression for the existence probability f is as follows:
[0032]
[0033] In the formula, P(M α ) represents road M in the simulation map. α Road feature sequence, P(X) α ) represents road X in the basic road data. α The road feature sequence, f represents the simulated map π. α The probability of the existence of a road feature sequence in the basic road data is denoted by m, where m represents the number of road feature sequences.
[0034] A computer system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for automatically generating static maps in a simulation test environment as described above.
[0035] A computer-readable storage medium having a computer program stored thereon, wherein when executed by a processor, the computer program implements the steps of the method for automatically generating static maps in a simulation test environment as described above.
[0036] The beneficial effects of this invention are as follows:
[0037] This invention utilizes real basic road data for sampling, and combines resampling and Markov properties to construct complex and realistic road features, thereby generating a simulated map topology with real road features.
[0038] This invention utilizes the Monte Carlo method to automatically generate simulated road maps on a large scale, thereby accelerating the iterative testing of autonomous driving algorithms and ensuring that the generated road topology has sufficient coverage and realism. Attached Figure Description
[0039] Figure 1 This is a flowchart of a method for automatically generating static maps in a simulation testing environment according to the present invention.
[0040] Figure 2 This is a schematic diagram of the data flow of a static map automatic generation method in a simulation test environment according to the present invention. Detailed Implementation
[0041] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.
[0042] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0043] To address the issues of existing technologies failing to generate large-scale simulation maps and producing inaccurate road topologies, this embodiment proposes an automatic static map generation method for simulation testing environments, such as... Figure 1 As shown, the method includes the following steps:
[0044] The real basic road data is divided into several first-class roads, and the feature sets of the first-class roads corresponding to the first-class roads are extracted;
[0045] A Gaussian mixture model is used to cluster several first road feature sets into several second road feature sets, wherein the second road feature sets include several first road feature sets;
[0046] The first road feature set in the second road feature set is resampled, and then the expected value of the first road feature set is calculated. The first road feature set with an expected value greater than a set threshold is selected to generate the road starting point.
[0047] By statistically analyzing the Markov properties of road features in real basic road data, the transition probabilities between various road features are constructed, and a simulated map of roads is generated by combining the road starting points.
[0048] The Monte Carlo method is used to evaluate the existence probability f of the roads in the generated simulation map, and road starting points are generated on a large scale to achieve the final generation of roads in the simulation map.
[0049] In this embodiment, the real basic road data can be obtained using real road map data or collected high-precision map data.
[0050] like Figure 2As shown, this embodiment slices the basic road data, dividing it into several first-class roads, including urban roads, highways, and expressways, and extracts the first-class road feature sets corresponding to urban roads, highways, and expressways respectively. Within the segmented first-class road map data, a Gaussian mixture model is used to cluster the extracted roads into second-class road feature sets, including mountain roads and plain roads. In this embodiment, the mountain roads and plain roads are the second-class roads and also the parent of the first-class roads; that is, mountain roads include urban roads, highways, and expressways, and similarly, plain roads also include urban roads, highways, and expressways. This classification is used because the road features corresponding to mountain roads and plain roads are not consistent.
[0051] In this embodiment, it is assumed that there are two sets of second road feature sets, each set of second road feature sets including three sets of first road feature sets, for a total of 6 sets of road feature sets. Based on specific practical needs, one set of first road feature sets from one set of second road feature sets is selected for resampling, and the expected value of the corresponding first road feature set is calculated. The first road feature set with an expected value greater than a set threshold is selected to generate the road starting point.
[0052] This invention utilizes real basic road data for sampling, and combines resampling and Markov properties to construct complex and realistic road features, thereby generating a simulated map topology with real road features.
[0053] This invention utilizes the Monte Carlo method to automatically generate simulated road maps on a large scale, thereby accelerating the iterative testing of autonomous driving algorithms and ensuring that the generated road topology has sufficient coverage and realism.
[0054] In this embodiment, the road features include road curvature, lateral slope, longitudinal slope, number of lanes, whether it is an intersection, intersection shape, lane line shape, lane line color, tunnels, and bridges.
[0055] In this embodiment, the features of the basic road data are labeled. The labeled road features cover the road features shown in Table 1. The specific parameters and symbols of each road feature are shown in Table 1.
[0056] Table 1 Road Characteristic Parameters
[0057]
[0058] In this embodiment, considering the large number of road feature parameters and the nonlinear and non-obvious coupling relationship between them, a Gaussian mixture model (GMM) with good fitting ability is selected for clustering. The calculation formula of the Gaussian mixture model is as follows:
[0059]
[0060] Where Θ=[K,ω k ,θ k K represents the number of Gaussian kernels in the Gaussian mixture model, which is also the number of first-class roads; ω k This represents the weight of the k-th Gaussian component in the Gaussian mixture model, and the sum of all weights equals 1. θ k =[μ k ,∑ k ] represents the mean matrix corresponding to the k-th Gaussian component ∑ k sum of variance matrix μ k ;x GMM Represents the road features of the input:
[0061] x GMM =[ρ,θ lat ,θ lon ,η,I,R,S,C,T,B] T (2)
[0062] Where ρ represents the road curvature, θ lat θ represents the lateral slope. lon The values represent the longitudinal slope, η represents the number of lanes, I represents whether it is an intersection, R represents the intersection shape, S represents the lane line shape, C represents the lane line color, T represents a tunnel, and B represents a bridge.
[0063] The number of Gaussian kernels in the Gaussian mixture model described in equation (1) is calculated as follows;
[0064] K min =arg min K (Kln(n c )-2ln(log(f(x GMM ;Θ)))) (3)
[0065] Where, arg min K This indicates finding the minimum value, where K represents the number of Gaussian kernels in the Gaussian mixture model, and n... c x represents the amount of data used for clustering. GMM K represents the input road features. min This represents the number of minimum Gaussian kernels obtained by solving the problem.
[0066] In this embodiment, real basic road data is used to perform statistical analysis according to the road features shown in Table 1. The feature distribution of each road feature parameter can be obtained based on the road feature parameters. In order to reduce the amount of typical high-probability maps and maps with low effectiveness in simulation testing, considering that there is a lot of duplicate data in the road feature set, a resampling mechanism is introduced to resample the road feature set. This is more conducive to constructing the starting points of road segments with different complex road features, and achieving more efficient and richer road feature coverage.
[0067] In this embodiment, for the first road feature set after resampling, the expected value of the first road feature set is calculated, and the first road feature set with an expected value greater than a set threshold is selected to generate the road starting point:
[0068]
[0069] Where, x i =[ρ,θ lat ,θ lon ,η,I,R,S,C,T,B] T Let ρ represent the road feature parameters sampled in the i-th time, and θ represent the road curvature. lat θ represents the lateral slope. lon The values represent the longitudinal slope, η represents the number of lanes, I represents whether it is an intersection, R represents the intersection shape, S represents the lane line shape, C represents the lane line color, T represents a tunnel, and B represents a bridge. x represents the sampled value i The value is given by n, where n represents the number of samples and E represents the expected value.
[0070] when When the expected value is greater than the set threshold, select the corresponding x. i Generate the starting point of the road.
[0071] In this embodiment, the generation of road geometry requires two conditions: first, the starting point of the road; and second, the road feature transition probabilities constructed based on real basic road data. Markov chains describe state transitions and are generally used to describe the dynamic transition characteristics of a system, while general stochastic processes describe the relationship between states and observation sequences. In this embodiment, the driving process is considered to consist of driving modes (hidden states) and observed states, and a Markov chain is used to describe the driving process. The initial state of the Markov chain is the road starting point x. i .
[0072] In this embodiment, the transition probabilities between various road features are constructed by statistically analyzing the Markov properties of road features in real basic road data, as shown below:
[0073]
[0074] Where, π α,β M represents the number of times road feature α is transferred to road feature β. α Represents road feature α;
[0075] Therefore, the state transition probability matrix π = [π] is obtained. α,β ] N*N , where N represents the size of the matrix.
[0076] The next road feature of a continuous road element can be obtained based on the magnitude of the transition probability. Therefore, by combining the transition probabilities between each road feature with the road starting point, a simulated road with realistic road characteristics is constructed, as shown in the following expression:
[0077] P(M α )=P(π α,β ,x i ), i = 1, 2, ..., n (6)
[0078] Among them, M α This refers to the roads in the generated simulation map; P(M) α ) represents road M in the simulation map. α The road feature sequence, x i Let P represent the starting point of the road and P represent the probability.
[0079] In order to ensure that the generated simulation map has sufficient coverage, this embodiment also uses the Monte Carlo method to evaluate the existence probability f of the generated simulation map roads. By controlling the existence probability f, the effectiveness of the generated simulation map roads can be selected.
[0080] The expression for the existence probability f is as follows:
[0081]
[0082] In the formula, P(M α ) represents road M in the simulation map. α Road feature sequence, P(X) α ) represents road X in the basic road data. α The road feature sequence, f represents the simulated map π. α The probability of the existence of a road feature sequence in the basic road data is denoted by m, where m represents the number of road feature sequences.
[0083] In this embodiment, based on real road data, the Monte Carlo method is used to automatically generate road starting points on a large scale for simulated road maps. Combined with resampling technology and Markov chain technology, complex and realistic road features are constructed, thereby accelerating the iterative testing of autonomous driving algorithms.
[0084] Because the initial state of the roads in the simulation map can theoretically have countless possibilities, but not every possible outcome is reasonable, Monte Carlo methods are used to evaluate their probabilities and select the most probable ones. Furthermore, Monte Carlo sampling generates a large number of initial states, resulting in a vast number of starting points.
[0085] In one specific embodiment, a computer system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method for automatically generating static maps in a simulation test environment as described above.
[0086] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.
[0087] In one specific embodiment, a computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the method for automatically generating static maps in a simulation test environment as described above.
[0088] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0089] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the claims of the present invention.
Claims
1. A method for automatically generating static maps in a simulation test environment, characterized in that: The method includes the following steps: The real basic road data is divided into several first-class roads, and the feature sets of the first-class roads corresponding to the first-class roads are extracted; A Gaussian mixture model is used to cluster several first road feature sets into several second road feature sets, wherein the second road feature sets include several first road feature sets; The first road feature set in the second road feature set is resampled, and then the expected value of the first road feature set is calculated. The first road feature set with an expected value greater than a set threshold is selected to generate the road starting point. By statistically analyzing the Markov properties of road features in real basic road data, the transition probabilities between various road features are constructed, and a simulated map of roads is generated by combining the road starting points. The Monte Carlo method was used to determine the probability of road existence in the generated simulation map. An evaluation was conducted, and road starting points were generated on a large scale to ultimately generate a simulated road map.
2. The method for automatically generating static maps in a simulation test environment according to claim 1, characterized in that: The road features mentioned include road curvature, lateral slope, longitudinal slope, number of lanes, whether it is an intersection, intersection shape, lane line shape, lane line color, tunnels, and bridges.
3. The method for automatically generating static maps in a simulation test environment according to claim 1, characterized in that: The calculation formula for the Gaussian mixture model is as follows: in, , This represents the number of Gaussian kernels in the Gaussian mixture model, which is also the number of Class I roads; This represents the weight of the k-th Gaussian component in the Gaussian mixture model, and the sum of all weights equals 1. ; This represents the mean matrix corresponding to the k-th Gaussian component. sum and variance matrix ; Represents the road features of the input: in, Indicates the curvature of the road. Indicates the lateral slope. Indicates longitudinal slope. Indicates the number of lanes. Indicates whether it is an intersection. Indicates the shape of the intersection. Indicates the shape of lane lines. Indicates the lane line color, Indicates a tunnel. It refers to a bridge.
4. The method for automatically generating static maps in a simulation test environment according to claim 3, characterized in that: The number of Gaussian kernels in the Gaussian mixture model is calculated as follows; in, This indicates finding the minimum value. This indicates the number of Gaussian kernels in the Gaussian mixture model. Indicates the amount of data used for clustering. This represents the input road features. This represents the number of minimum Gaussian kernels obtained by solving the problem.
5. The method for automatically generating static maps in a simulation test environment according to claim 1, characterized in that: The formula for calculating the starting point of a road is as follows: in, Indicates the first i Secondary sampling of road feature parameters Indicates the curvature of the road. Indicates the lateral slope. Indicates longitudinal slope. Indicates the number of lanes. Indicates whether it is an intersection. Indicates the shape of the intersection. Indicates the shape of lane lines. Indicates the lane line color, Indicates a tunnel. Indicates a bridge. Indicates sampling , Indicates the number of samples. This indicates a demand for expectation; when When the value is greater than the set threshold, select the corresponding option. Generate the starting point of the road.
6. The method for automatically generating static maps in a simulation test environment according to claim 1, characterized in that: The transition probabilities between various road features are constructed as follows: in, Indicates road characteristics Transfer to road features Number of times, Indicates road characteristics ; This yields the state transition probability matrix. ,in, Indicates the size of the matrix.
7. The method for automatically generating static maps in a simulation test environment according to claim 6, characterized in that: By combining the transition probabilities between various road features with the road starting point, a simulated road with realistic road characteristics is constructed, as shown in the following expression: in, This refers to the generated simulated map roads; Represents roads in a simulation map The road feature sequence, Indicates the starting point of the road. It represents probability.
8. The method for automatically generating static maps in a simulation test environment according to claim 1, characterized in that: The probability of existence The expression is as follows: In the formula, Represents roads in a simulation map The road feature sequence, Represents roads in basic road data The road feature sequence, Represents a simulation map The probability of the existence of road feature sequences in the basic road data. This indicates the number of road feature sequences.
9. A computer system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for automatically generating static maps in a simulation test environment as described in any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method for automatically generating static maps in a simulation test environment as described in any one of claims 1 to 8.