Monte Carlo simulation-based diversified risk level trajectory simulation method and system

Abstract

The invention discloses a diversified risk level trajectory simulation method and system based on Monte Carlo simulation. The method comprises the steps that low-speed yielding trajectory segment data and initial acceleration trajectory segment data of a target interaction vehicle are extracted according to high-interaction vehicle trajectory data; carrying out data distribution fitting to obtain an optimal distribution parameter of the way-giving time and an optimal distribution parameter of the initial acceleration; using Monte Carlo simulation to generate target duration time, and adjusting the frame number of the low-speed way-giving track segment data according to the target duration time to obtain a low-speed way-giving target track; using Monte Carlo simulation to generate a target initial acceleration, and using an MPC controller to track a reference path according to the target initial acceleration to obtain an initial acceleration section target trajectory; and combining the low-speed yielding target trajectory and the initial acceleration section target trajectory to obtain a target simulation trajectory, and determining a risk level of the target simulation trajectory. The method improves the efficiency and reliability of track simulation, and can be applied to the technical field of automatic driving.

Description

Technical Field

[0001] The present invention relates to the technical field of trajectory simulation, and in particular to a method and system for simulating trajectories with diverse risk levels based on Monte Carlo simulation. Background Art

[0002] By collecting and analyzing the motion trajectory data of a large number of real vehicles, statistical features reflecting real traffic behaviors can be extracted, including vehicle speed, acceleration, driving path, etc. The distribution characteristics of these data provide a basis for the simulator to generate real and reasonable traffic trajectories. Especially in high-interaction scenarios, capturing and fitting the distribution of the low-speed yielding time and initial acceleration of the target interacting vehicle is crucial for simulating different risk trajectories and evaluating collision risks. Based on this, how to generate simulated trajectories with diverse risk levels has become an urgent problem to be solved. Summary of the Invention

[0003] An object of the present invention is to solve at least to some extent one of the technical problems existing in the prior art.

[0004] To this end, an object of an embodiment of the present invention is to provide a method for simulating trajectories with diverse risk levels based on Monte Carlo simulation, which can generate vehicle simulated trajectories with diverse risk levels and improve the efficiency and reliability of trajectory simulation.

[0005] Another object of an embodiment of the present invention is to provide a system for simulating trajectories with diverse risk levels based on Monte Carlo simulation.

[0006] To achieve the above technical object, the technical solutions adopted in the embodiments of the present invention include:

[0007] In a first aspect, an embodiment of the present invention provides a method for simulating trajectories with diverse risk levels based on Monte Carlo simulation, including the following steps:

[0008] Obtain high-interaction vehicle trajectory data, and extract the low-speed yielding trajectory segment data and initial acceleration trajectory segment data of the target interacting vehicle according to the high-interaction vehicle trajectory data;

[0009] Perform data distribution fitting on the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data respectively to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration;

[0010] According to the low-speed yielding trajectory segment data and the optimal distribution parameters of the yielding time, use Monte Carlo simulation to generate a target duration, and adjust the number of frames of the low-speed yielding trajectory segment data according to the target duration to obtain a low-speed yielding target trajectory;

[0011] According to the initial acceleration trajectory segment data and the initial acceleration optimal distribution parameters, use Monte Carlo simulation to generate a target initial acceleration, and use an MPC controller to track a reference path according to the target initial acceleration to obtain an initial acceleration segment target trajectory;

[0012] Merge the low-speed yielding target trajectory and the initial acceleration segment target trajectory to obtain a target simulation trajectory, and determine the risk level of the target simulation trajectory.

[0013] Further, in an embodiment of the present invention, the obtaining of the high-interaction vehicle trajectory data and the extraction of the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data of the target interaction vehicle according to the high-interaction vehicle trajectory data specifically include:

[0014] Obtain the trajectory data of a target interaction vehicle passing through an intersection and whose trajectory path is crossed by surrounding vehicles to obtain initial trajectory data;

[0015] Perform data cleaning and trajectory identification on the initial trajectory data to obtain the high-interaction vehicle trajectory data;

[0016] Extract the low-speed yielding trajectory frame segments of the target interaction vehicle with a speed lower than a preset speed threshold according to the high-interaction vehicle trajectory data, and determine the corresponding number of continuous frames and duration to obtain the low-speed yielding trajectory segment data;

[0017] Extract the initial acceleration trajectory frame segments of the target interaction vehicle whose speed first reaches the preset speed threshold according to the high-interaction vehicle trajectory data, and determine the corresponding initial acceleration to obtain the initial acceleration trajectory segment data.

[0018] Further, in an embodiment of the present invention, the respectively performing data distribution fitting on the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration specifically include:

[0019] Use a lognormal distribution and a gamma distribution to perform data distribution fitting on the low-speed yielding trajectory segment data, and determine the optimal distribution type and the corresponding fitting parameters based on a preset evaluation index to obtain the optimal distribution parameters of the yielding time;

[0020] Use a lognormal distribution and a gamma distribution to perform data distribution fitting on the initial acceleration trajectory segment data, and determine the optimal distribution type and the corresponding fitting parameters based on a preset evaluation index to obtain the optimal distribution parameters of the initial acceleration;

[0021] Among them, if the optimal distribution type is the lognormal distribution, the corresponding fitting parameters include the logarithmic mean and the logarithmic standard deviation. If the optimal distribution type is the gamma distribution, the corresponding fitting parameters include the shape parameter and the scale parameter.

[0022] Further, in an embodiment of the present invention, the method for generating a target duration by using Monte Carlo simulation according to the low-speed yielding trajectory segment data and the optimal distribution parameters of the yielding time, and adjusting the number of frames of the low-speed yielding trajectory segment data according to the target duration to obtain a low-speed yielding target trajectory specifically includes:

[0023] Randomly sample the low-speed yielding trajectory segment data by using Monte Carlo simulation according to the optimal distribution parameters of the yielding time, and statistically obtain the target duration according to the sampling result;

[0024] When the duration of the low-speed yielding trajectory frame segment is less than the target duration, perform interpolation processing on the low-speed yielding trajectory frame segment to obtain the adjusted low-speed yielding trajectory frame segment;

[0025] When the duration of the low-speed yielding trajectory frame segment is greater than the target duration, perform cropping processing on the low-speed yielding trajectory frame segment to obtain the adjusted low-speed yielding trajectory frame segment;

[0026] Perform distance correction on the adjusted low-speed yielding trajectory frame segment to obtain the low-speed yielding target trajectory.

[0027] Further, in an embodiment of the present invention, the method for generating a target initial acceleration by using Monte Carlo simulation according to the initial acceleration trajectory segment data and the optimal distribution parameters of the initial acceleration, and tracking a reference path by using an MPC controller according to the target initial acceleration to obtain an initial acceleration segment target trajectory specifically includes:

[0028] Randomly sample the initial acceleration trajectory segment data by using Monte Carlo simulation according to the optimal distribution parameters of the initial acceleration, and statistically obtain the target initial acceleration according to the sampling result;

[0029] Determine an acceleration start frame according to the initial acceleration trajectory frame segment, and determine the initial state of the vehicle corresponding to the acceleration start frame;

[0030] Optimize and update the vehicle state of the target interaction vehicle by using an MPC controller according to the target initial acceleration and the vehicle initial state to generate a simulated trajectory frame;

[0031] Obtain the low-speed yielding target trajectory according to the acceleration start frame and the simulated trajectory frame.

[0032] Further, in an embodiment of the present invention, the merging of the low-speed yielding target trajectory and the initial acceleration segment target trajectory to obtain a target simulation trajectory specifically includes:

[0033] Obtain the low-speed yielding target trajectory and the initial acceleration segment target trajectory corresponding to the same target interaction vehicle;

[0034] Determine the ending frame number of the initial acceleration segment target trajectory, and re-number the trajectory frames of the low-speed yielding target trajectory according to the ending frame number, and then add the re-numbered low-speed yielding target trajectory after the initial acceleration segment target trajectory to obtain the target simulation trajectory.

[0035] Further, in an embodiment of the present invention, the determination of the risk level of the target simulation trajectory specifically includes:

[0036] Determine the vehicle danger level of the target interaction vehicle based on the post-encroachment time;

[0037] Determine the risk perception ability of the target interaction vehicle based on the driving risk field;

[0038] Comprehensively evaluate the risk level of the target simulation trajectory according to the vehicle danger level and the risk perception ability.

[0039] In a second aspect, an embodiment of the present invention provides a diversified risk level trajectory simulation system based on Monte Carlo simulation, including:

[0040] A trajectory data extraction module, configured to obtain high-interaction vehicle trajectory data, and extract low-speed yielding trajectory segment data and initial acceleration trajectory segment data of a target interaction vehicle according to the high-interaction vehicle trajectory data;

[0041] A data distribution fitting module, configured to respectively perform data distribution fitting on the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data to obtain an optimal distribution parameter for yielding time and an optimal distribution parameter for initial acceleration;

[0042] A low-speed yielding target trajectory generation module, configured to generate a target duration by using Monte Carlo simulation according to the low-speed yielding trajectory segment data and the optimal distribution parameter for yielding time, and adjust the number of frames of the low-speed yielding trajectory segment data according to the target duration to obtain a low-speed yielding target trajectory;

[0043] An initial acceleration segment target trajectory generation module, configured to generate a target initial acceleration by using Monte Carlo simulation according to the initial acceleration trajectory segment data and the optimal distribution parameter for initial acceleration, and use an MPC controller to track a reference path according to the target initial acceleration to obtain an initial acceleration segment target trajectory;

[0044] A trajectory merging module, configured to merge the low-speed yielding target trajectory and the initial acceleration segment target trajectory to obtain a target simulation trajectory, and determine the risk level of the target simulation trajectory.

[0045] In a third aspect, an embodiment of the present invention provides a diversified risk level trajectory simulation device based on Monte Carlo simulation, including:

[0046] At least one processor;

[0047] At least one memory, configured to store at least one program;

[0048] When the at least one program is executed by the at least one processor, the at least one processor is caused to implement the above-mentioned diversified risk level trajectory simulation method based on Monte Carlo simulation.

[0049] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, in which a processor-executable program is stored, and the processor-executable program is used to execute the above-mentioned diversified risk level trajectory simulation method based on Monte Carlo simulation when executed by a processor.

[0050] The advantages and beneficial effects of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention:

[0051] In an embodiment of the present invention, high-interaction vehicle trajectory data is obtained. Based on the high-interaction vehicle trajectory data, the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data of the target interaction vehicle are extracted. The data distribution fitting is respectively performed on the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration. According to the low-speed yielding trajectory segment data and the optimal distribution parameters of the yielding time, the target duration is generated by using Monte Carlo simulation, and the number of frames of the low-speed yielding trajectory segment data is adjusted according to the target duration to obtain the low-speed yielding target trajectory. According to the initial acceleration trajectory segment data and the optimal distribution parameters of the initial acceleration, the target initial acceleration is generated by using Monte Carlo simulation, and the reference path is tracked by using an MPC controller according to the target initial acceleration to obtain the initial acceleration segment target trajectory. The low-speed yielding target trajectory and the initial acceleration segment target trajectory are merged to obtain the target simulation trajectory, and the risk level of the target simulation trajectory is determined. In the embodiment of the present invention, the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data of the target interaction vehicle are extracted based on the high-interaction vehicle trajectory data, and the data distribution fitting is respectively performed to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration. The target duration and the target initial acceleration are generated by using Monte Carlo simulation according to the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration, so as to adjust the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data, obtain the low-speed yielding target trajectory and the initial acceleration segment target trajectory, and then merge them to obtain the target simulation trajectory, which can generate vehicle simulation trajectories with diverse risk levels and improve the efficiency and reliability of trajectory simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following introduces the drawings required to be used in the embodiments of the present invention. It should be understood that the drawings introduced below are only for conveniently and clearly presenting some embodiments of the technical solutions in the present invention, and those skilled in the art can also obtain other drawings based on these drawings without creative efforts.

[0053] Figure 1 It is a flowchart of the steps of a method for simulating trajectories with diverse risk levels based on Monte Carlo simulation provided by an embodiment of the present invention;

[0054] Figure 2 It is a schematic diagram of the distribution fitting of the yielding time provided by an embodiment of the present invention;

[0055] Figure 3 It is a schematic diagram of the distribution fitting of the initial acceleration provided by an embodiment of the present invention;

[0056] Figure 4 It is a three-dimensional spatio-temporal diagram of the low-speed yielding trajectory segment generated by simulation provided by an embodiment of the present invention;

[0057] Figure 5 Schematic diagram of the initially accelerated trajectory segment generated by simulation provided by an embodiment of the present invention;

[0058] Figure 6 Schematic diagram of the distribution of the minimum PET value of the target simulation trajectory provided by an embodiment of the present invention;

[0059] Figure 7 Schematic diagram of the cumulative value of PET≤2s of the target simulation trajectory provided by an embodiment of the present invention;

[0060] Figure 8 Schematic diagram of the distribution of the DRF value of the target simulation trajectory provided by an embodiment of the present invention;

[0061] Figure 9 Block diagram of the structure of a diversified risk level trajectory simulation system based on Monte Carlo simulation provided by an embodiment of the present invention;

[0062] Figure 10 Block diagram of the structure of a diversified risk level trajectory simulation device based on Monte Carlo simulation provided by an embodiment of the present invention. Detailed implementation manners

[0063] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary only for explaining the present invention and should not be construed as limiting the present invention. For the step numbers in the following embodiments, they are only set for the convenience of description and explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0064] In the description of the present invention, the meaning of "a plurality of" is two or more. If there is a description of "first" and "second", it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence of the indicated technical features. In addition, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field of the present invention.

[0065] Referring to Figure 1 , an embodiment of the present invention provides a diversified risk level trajectory simulation method based on Monte Carlo simulation, which specifically includes the following steps:

[0066] S101. Obtain high-interaction vehicle trajectory data, and extract the low-speed yielding trajectory segment data and initially accelerated trajectory segment data of the target interaction vehicle according to the high-interaction vehicle trajectory data;

[0067] S102. Respectively perform data distribution fitting on the slowdown yielding trajectory segment data and the initial acceleration trajectory segment data to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration.

[0068] S103. According to the slowdown yielding trajectory segment data and the optimal distribution parameters of the yielding time, use Monte Carlo simulation to generate a target duration, and adjust the number of frames of the slowdown yielding trajectory segment data according to the target duration to obtain the slowdown yielding target trajectory.

[0069] S104. According to the initial acceleration trajectory segment data and the optimal distribution parameters of the initial acceleration, use Monte Carlo simulation to generate a target initial acceleration, and use an MPC controller to track a reference path according to the target initial acceleration to obtain the initial acceleration segment target trajectory.

[0070] S105. Merge the slowdown yielding target trajectory and the initial acceleration segment target trajectory to obtain a target simulation trajectory, and determine the risk level of the target simulation trajectory.

[0071] In the embodiment of the present invention, the slowdown yielding trajectory segment data and the initial acceleration trajectory segment data of the target interaction vehicle are extracted from the high-interaction vehicle trajectory data, and data distribution fitting is respectively performed to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration. According to the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration, Monte Carlo simulation is used to generate a target duration and a target initial acceleration, so as to adjust the slowdown yielding trajectory segment data and the initial acceleration trajectory segment data, obtain the slowdown yielding target trajectory and the initial acceleration segment target trajectory, and then merge them to obtain a target simulation trajectory, which can generate vehicle simulation trajectories with diverse risk levels, improving the efficiency and reliability of trajectory simulation.

[0072] Further as an optional implementation manner, high-interaction vehicle trajectory data is obtained, and the slowdown yielding trajectory segment data and the initial acceleration trajectory segment data of the target interaction vehicle are extracted from the high-interaction vehicle trajectory data, which specifically includes:

[0073] S1011. Obtain the trajectory data of the target interaction vehicle passing through the intersection and whose trajectory path is crossed by surrounding vehicles to obtain initial trajectory data.

[0074] S1012. Perform data cleaning and trajectory identification on the initial trajectory data to obtain high-interaction vehicle trajectory data.

[0075] S1013. Extract the slowdown yielding trajectory frame segments of the target interaction vehicle with a speed lower than a preset speed threshold from the high-interaction vehicle trajectory data, and determine the corresponding number of continuous frames and duration to obtain the slowdown yielding trajectory segment data.

[0076] S1014. Extract the initial acceleration trajectory frame segment of the target interaction vehicle whose speed first reaches the preset speed threshold from the high-interaction vehicle trajectory data, and determine the corresponding initial acceleration to obtain the initial acceleration trajectory segment data.

[0077] Specifically, the data used in the embodiments of the present invention is derived from the trajectories recorded from the perspective of an unmanned aerial vehicle at a real urban un-signalized intersection. The dataset includes traffic flow information during multiple recording periods, specifically including the following types of data:

[0078] recordingId: Represents different acquisition periods

[0079] trackId: The unique identifier of the vehicle during this period, used to distinguish different vehicles.

[0080] frame: The video frame number, used to represent the time series, and each frame represents a time interval of 0.4 seconds.

[0081] xCenter, yCenter: The central position coordinates of the vehicle on the two-dimensional plane, with the unit of meter (m).

[0082] xVelocity, yVelocity: The velocity components of the vehicle in the x and y directions, with the unit of meter per second (m / s).

[0083] xAcceleration, yAcceleration: The acceleration components of the vehicle in the x and y directions, with the unit of meter per second squared (m / s 2 )

[0084] ms: The time identifier, used to mark critical moments.

[0085] Width: The vehicle length.

[0086] Length: The vehicle width.

[0087] The highly interactive trajectory is defined as the trajectory path of the target interaction vehicle (i.e., the vehicle to be turned) being crossed by surrounding vehicles, and the target interaction vehicle passes through the intersection.

[0088] Data cleaning includes physical parameter inspection and data integrity inspection. Physical parameter inspection mainly excludes the trajectory data with the vehicle width (width) or length (length) being zero. These abnormal data are often detection noise points or extreme outliers, which may have an adverse impact on subsequent analysis. Data integrity inspection ensures that the key fields (such as position, speed, acceleration) in the trajectory data are not empty and are reasonable values, preventing missing values or outliers from interfering with the data analysis results.

[0089] To facilitate subsequent data processing, recordingId and trackId are combined to generate a unique identifier (unique_id) to ensure that the track of each vehicle can be uniquely identified and tracked in the data set after multiple files are merged.

[0090] Setting critical speed threshold When the vehicle speed is lower than this threshold, the vehicle is considered to be in a low-speed yielding state. The basis for selecting this threshold is that in actual intersection vehicle crossing behavior, the yielding behavior of vehicles before the stop line is usually accompanied by a significant drop in speed. Vehicles below this speed are more likely to be in a state of observing or waiting for the opportunity to pass.

[0091] In each vehicle trajectory to be turned, the vehicle speed v in consecutive frames is identified <v threshold For each trajectory, sort it by time sequence (frame number) to ensure that the time sequence of the frame segments is correct. Use Boolean index to identify whether each frame meets the low speed condition, that is, v i <v threshold By detecting the change of speed mark, the frame segments that continuously meet the low-speed condition are divided. The starting frame s and the ending frame e of each low-speed frame segment are recorded, that is, the frame segment [s, e] meets (v i <v threshold )for i=s,s+1,…,e.

[0092] For each identified low-speed frame segment, its number of frames (n=e-s+1) and duration (t=n×Δt) are calculated and saved, where Δt=0.4s is the frame interval time.

[0093] Get the coordinates of the stop line, get the corresponding frame of the vehicle to be turned near the stop line through the metadata trajectory file, and record the first time the speed exceeds the critical speed threshold The corresponding frame is marked as the initial acceleration keyframe.

[0094] The initial acceleration a calculated for all vehicles at the key frame time i The sample set {a1, a2, ..., a M}, where M is the number of samples. Data cleaning is performed during this process to ensure that all acceleration values ​​are non-negative and valid, and to remove any abnormal or missing data.

[0095] As an optional implementation, data distribution fitting is performed on the low-speed yield trajectory segment data and the initial acceleration trajectory segment data to obtain the optimal distribution parameters of the yield time and the optimal distribution parameters of the initial acceleration, which specifically include:

[0096] S1021. Fit the data distribution of the slow-down and give-way trajectory segments using the lognormal distribution and the gamma distribution, determine the optimal distribution type and the corresponding fitting parameters based on a preset evaluation index, and obtain the optimal distribution parameters of the give-way time.

[0097] S1022. Fit the data distribution of the initial acceleration trajectory segments using the lognormal distribution and the gamma distribution, determine the optimal distribution type and the corresponding fitting parameters based on a preset evaluation index, and obtain the optimal distribution parameters of the initial acceleration.

[0098] Among them, if the optimal distribution type is the lognormal distribution, the corresponding fitting parameters include the log mean and the log standard deviation; if the optimal distribution type is the gamma distribution, the corresponding fitting parameters include the shape parameter and the scale parameter.

[0099] Specifically, in order to accurately describe the random characteristics of the slow-down and give-way duration and the initial acceleration of the highly interactive trajectories in the original dataset, based on the right-skewed characteristics of the original data, this section selects the lognormal distribution (Lognormal) and the gamma distribution (Gamma) as fitting models and determines the optimal distribution through statistical methods.

[0100] The lognormal distribution is suitable for describing non-negative random variables whose logarithmic values follow a normal distribution. In traffic behavior analysis, many time intervals (such as reaction time, waiting time) often exhibit lognormal distribution characteristics. Its probability density function (PDF) is:

[0101]

[0102] x represents the slow-down and give-way segment duration or the initial acceleration of the initial acceleration segment of the vehicle waiting to turn. μ is the log mean, representing the mean of lnx, and σ is the log standard deviation, representing the standard deviation of lnx.

[0103] Parameter estimation is solved by the maximum likelihood estimation method, minimizing the negative log-likelihood function to obtain μ and σ.

[0104] The gamma distribution is a flexible continuous distribution suitable for describing non-negative random variables, especially those with variable shape parameters. Its probability density function (PDF) is:

[0105]

[0106] x represents the slow-down and give-way segment duration or the initial acceleration of the initial acceleration segment of the vehicle waiting to turn. α is the shape parameter, controlling the shape of the distribution; β is the scale function, affecting the extent of the distribution; Γ(α) is the gamma function, defined as

[0107] Similarly, parameter estimation uses the maximum likelihood estimation method to minimize the negative log-likelihood function to solve for α and β.

[0108] After completing the parameter estimation of the log-normal distribution and the gamma distribution, it is necessary to evaluate the goodness of fit of the model through various statistical indicators, and then select the optimal distribution model. Specifically, it includes the Log-Likelihood, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Kolmogorov-Smirnov test (K-S Test).

[0109] The log-likelihood function measures the ability of the model to explain the observed data under given parameters and is defined as:

[0110]

[0111] where θ represents the model parameters (such as (μ,σ) or (α,β)), and f(t i ; θ) is the probability density function of the corresponding distribution. The larger the log-likelihood value, the stronger the ability of the model to explain the data.

[0112] AIC combines the goodness of fit and complexity of the model, and its calculation formula is:

[0113]

[0114] where k is the number of model parameters, is the maximized log-likelihood value. The lower the AIC, the better the model, because while considering the goodness of fit of the model, it penalizes the complexity of the model to prevent overfitting.

[0115] The calculation formula of BIC is as follows:

[0116]

[0117] where N is the number of samples. Similar to AIC, BIC also tends to select a lower value, indicating a better model. Compared with AIC, BIC penalizes the complexity of the model more strictly, especially when the sample size is large.

[0118] The K-S test is used to compare the cumulative distribution function (CDF) of the sample data with the CDF of the theoretical distribution, and test the null hypothesis (H0): the sample comes from the specified distribution. Calculate the K-S statistic (D) and the corresponding p-value:

[0119]

[0120] Among them, sup represents taking the maximum value, and F empirical (t) is the empirical cumulative distribution function of the sample data, and F theoretical (t) is the cumulative distribution function of the theoretical distribution. The larger the p-value, the more it indicates that H0 cannot be rejected, that is, the theoretical distribution has a higher degree of agreement with the sample data.

[0121] Taking into account the results of AIC, BIC, and K-S tests comprehensively, preferentially select the models with lower AIC and BIC values. When the indicators are similar, consider the p-value of the K-S test and select the model with a higher p-value.

[0122] After determining the optimal distribution model through the above evaluation steps, save the distribution type and parameter values of the model to a pickle file.

[0123] Use the kernel density estimation method (the kdeplot function of Seaborn) to plot the probability density fitting curves of the yielding time in the low-speed yielding trajectory segment and the initial acceleration in the initial acceleration segment of the vehicle to be turned.

[0124] Figure 2 and Figure 3 respectively show the schematic diagrams of the fitting curves of the yielding time in the low-speed yielding trajectory segment and the initial acceleration in the initial acceleration trajectory segment fitted by the gamma distribution and the lognormal distribution. Among them, Figure 2 and Figure 3 The vertical axis of represents the probability density, Figure 2 and Figure 3 The horizontal axes of and represent the yielding time and the initial acceleration respectively.

[0125] Furthermore, as an optional implementation method, according to the data of the low-speed yielding trajectory segment and the optimal distribution parameters of the yielding time, use Monte Carlo simulation to generate the target duration, and adjust the number of frames of the low-speed yielding trajectory segment data according to the target duration to obtain the low-speed yielding target trajectory, which specifically includes:

[0126] S1031. According to the optimal distribution parameters of the yielding time, use Monte Carlo simulation to randomly sample the data of the low-speed yielding trajectory segment, and statistically obtain the target duration according to the sampling results;

[0127] S1032. When the duration of the low-speed yielding trajectory frame segment is less than the target duration, perform interpolation processing on the low-speed yielding trajectory frame segment to obtain the adjusted low-speed yielding trajectory frame segment;

[0128] S1033. When the duration of the low-speed yielding trajectory frame segment is greater than the target duration, perform cropping processing on the low-speed yielding trajectory frame segment to obtain the adjusted low-speed yielding trajectory frame segment;

[0129] S1034. Perform distance correction on the adjusted low-speed yielding trajectory frame segments to obtain the low-speed yielding target trajectory.

[0130] Specifically, the Monte Carlo method is a class of statistical methods that obtain numerical results through random sampling. Its core idea is that if we know that a target variable (such as the yielding duration) follows a probability distribution, we can obtain a large sample through repeated sampling, thus realistically reproducing the diversity and uncertainty of the variable in the real world in a random sense. In the first part of this study, the optimal distribution of the vehicle's low-speed yielding duration was fitted, and this distribution can be used as the basis for Monte Carlo sampling.

[0131] First, it is necessary to load the optimal distribution parameters fitted in the first part. These parameters include the distribution type (lognormal or gamma) and the corresponding shape parameter (shape), location parameter (loc), and scale parameter (scale). These parameters will serve as the basis for Monte Carlo simulation to generate duration samples that conform to the characteristics of real traffic behavior. At the same time, read all the low-speed yielding trajectory segment data. Each trajectory segment data contains the spatial coordinates and speed information of the vehicle in front of the stop line, which provides the necessary reference and benchmark for subsequent trajectory segment adjustment.

[0132] Based on the loaded optimal distribution parameters, use the Monte Carlo method to perform a large number of random samplings to generate the target duration. Specifically, assume that the optimal distribution is a lognormal distribution with parameters μ and σ, then the generation process of the target duration T new is as follows:

[0133] T new ~Lognormal(μ,σ)

[0134] Similarly, if the optimal distribution is a gamma distribution with parameters α and β, the generation process of the target duration is:

[0135] T new ~Γ(α,β)

[0136] Among them, μ represents the logarithmic mean, corresponding to the logarithmic mean of the lognormal distribution. σ represents the logarithmic standard deviation, corresponding to the logarithmic standard deviation of the lognormal distribution. α represents the shape parameter, corresponding to the shape parameter of the gamma distribution. β represents the scale parameter, corresponding to the scale parameter of the gamma distribution. The generated target duration T new will be used to adjust the number of frames of the original trajectory segment, so as to achieve the time extension or shortening of the trajectory segment.

[0137] According to the generated target duration T new , it is necessary to adjust the number of frames N of the original trajectory segment (the original duration T original=(N - 1)×Δt, where Δt = 0.4s is the frame interval time), to match the new duration T new .

[0138] When the duration of the low-speed yielding trajectory frame segment is less than the target duration, i.e., the target number of frames it is necessary to interpolate the trajectory segment and add several frames to extend the duration. First, calculate the number of frames to be inserted ΔN = N new - N. Then evenly distribute the inserted frames to each position of the trajectory segment to ensure the smoothness of the interpolation process. For each pair of adjacent trajectory points (x i , y i ) and (x i+1 , y i+1 ), according to the interpolation ratio (where k is the index of the inserted frame), calculate the coordinates of the newly inserted frame:

[0139] x new = x i + α(x i+1 - x i )

[0140] y new = y i + α(y i+1 - y i )

[0141] where the value range of α is 0 < α < 1 to ensure that the newly inserted frame is located between the original two points.

[0142] After interpolation, recalculate the speed v x and v y , as well as the acceleration a x and a y to ensure the physical consistency of the trajectory segment.

[0143] When the duration of the low-speed yielding trajectory frame segment is greater than the target duration, i.e., the target number of frames it is necessary to crop the trajectory segment and delete several frames to shorten the duration. First, calculate the number of frames to be deleted ΔN = N new - N. By deleting frames at equal intervals, avoid deleting frames within a certain time period concentratedly and maintain the overall coherence of the trajectory segment. Calculate the interval for deleting frames:

[0144]

[0145] Determine the index positions of the frames to be deleted and regenerate the adjusted trajectory segment:

[0146]

[0147] where Represents a floor operation to ensure that the index is within the valid range.

[0148] After cropping, recalculate the velocity v for each frame x With v y And acceleration a x With a y , to ensure the physical consistency of the trajectory segment. For each frame i (starting from the second frame), the velocity components are calculated from the position change and the time interval:

[0149]

[0150] Where x i And y i Represent the x and y coordinates of the i-th frame respectively, and Δt = 0.4s is the frame interval time.

[0151] For each frame i (starting from the second frame), the acceleration components are calculated from the position change and the time interval:

[0152]

[0153] Where v x,i And v y,i Represent the x and y velocity components of the i-th frame respectively, and Δt = 0.4s is the frame interval time.

[0154] During the trajectory segment adjustment process, the generated new trajectory segment may cause the changes in velocity and acceleration to exceed the reasonable range of vehicle dynamics due to interpolation or cropping operations. Therefore, a maximum threshold A of acceleration is set max = 3.0m / s 2 .

[0155] When the acceleration at any moment in the adjusted trajectory segment Exceeds this threshold A max , it is necessary to smooth the velocity to limit the amplitude of the acceleration. The window size is set to 3 frames, that is, the velocities of the current frame and the frames before and after it are averaged. For each frame i:

[0156]

[0157] Where And Represent the smoothed velocity components.

[0158] The smoothed velocity components are used to recalculate the acceleration to ensure that the acceleration amplitude does not exceed A max :

[0159]

[0160] Among them, Δt = 0.4 s is the frame interval time.

[0161] After the trajectory segment is adjusted, especially after the interpolation and cropping operations, the total driving distance D of the trajectory segment new may deviate from the total driving distance D of the original trajectory segment original There is a deviation. To maintain the spatial consistency of the trajectory, the distance correction is performed by the following method.

[0162] For the original trajectory segment, calculate the total distance D original :

[0163]

[0164] For the adjusted trajectory segment, calculate the total distance D new :

[0165]

[0166] To correct the total distance of the adjusted trajectory segment to be consistent with the original trajectory segment, calculate the distance proportionality coefficient λ:

[0167] (If D new ≠0 otherwise λ = 1.0)

[0168] Scale the speed components of the adjusted trajectory segment:

[0169]

[0170] Finally, its spatial movement distance is consistent with the original trajectory segment. As Figure 4 shown is the three-dimensional spatio-temporal diagram of the simulated low-speed yielding trajectory segment provided by the embodiment of the present invention.

[0171] Further as an optional implementation manner, according to the initial acceleration trajectory segment data and the initial acceleration optimal distribution parameters, use Monte Carlo simulation to generate the target initial acceleration, and use the MPC controller to track the reference path according to the target initial acceleration to obtain the target trajectory of the initial acceleration segment, which specifically includes:

[0172] S1041. According to the initial acceleration optimal distribution parameters, use Monte Carlo simulation to randomly sample the initial acceleration trajectory segment data, and statistically obtain the target initial acceleration according to the sampling results;

[0173] S1042. Determine the acceleration start frame according to the initial acceleration trajectory frame segment, and determine the vehicle initial state corresponding to the acceleration start frame;

[0174] S1043. According to the target initial acceleration and the vehicle initial state, use the MPC controller to optimize and update the vehicle state of the target interacting vehicle to generate the simulated trajectory frame;

[0175] S1044. Obtain the low-speed yielding target trajectory based on the acceleration start frame and the simulated trajectory frame.

[0176] Specifically, first, it is necessary to load the optimal distribution parameters obtained by the first part of the fitting. These parameters include the distribution type (lognormal or gamma) and the corresponding shape parameter (shape), location parameter (loc), and scale parameter (scale).

[0177] Based on the loaded optimal distribution parameters, use the Monte Carlo method to perform a large number of random samplings to generate the target initial acceleration. Specifically, assume that the optimal distribution is a lognormal distribution with parameters μ and σ, then the generation process of the target initial acceleration a i is as follows:

[0178] a i ~Lognormal(μ,σ)

[0179] Similarly, if the optimal distribution is a gamma distribution with parameters α and β, then the generation process of the target initial acceleration a i is:

[0180] a i ~Γ(α,β)

[0181] Among them, μ represents the logarithmic mean, corresponding to the logarithmic mean of the lognormal distribution. σ represents the logarithmic standard deviation, corresponding to the logarithmic standard deviation of the lognormal distribution. α represents the shape parameter, corresponding to the shape parameter of the gamma distribution. β represents the scale parameter, corresponding to the scale parameter of the gamma distribution. The generated target initial acceleration a i will be used as the input to the MPC controller to update the vehicle state and achieve the regeneration of the trajectory segment. To ensure physical rationality, the maximum acceleration limit is imposed on the sampling results. Referring to the maximum acceleration at which the true value appears, it is set to 4.1 m / s 2 .

[0182] By screening the marked key frames, determine the starting frame number (frame) when the vehicle officially enters the initial acceleration stage. This frame marks the transition of the vehicle from a stopped or low-speed state to a normal driving state. According to the starting frame number, extract the trajectory segment of the vehicle behind the stop line (frame >= initial_frame) as the reference path.

[0183] Calculate the heading angle ψ for each frame through the velocity component:

[0184]

[0185] Set the minimum number of frames of the reference path to ensure that the MPC controller has sufficient trajectory data for optimization within the prediction horizon.

[0186] Obtain the initial state x0, y0, v of the vehicle at the starting frame (x,0) , v (y,0) , a0, ψ0.

[0187] Use a simplified bicycle model to describe the dynamic behavior of the vehicle. The vehicle state vector x is defined as:

[0188]

[0189] x, y: The position coordinates of the vehicle in the two-dimensional plane;

[0190] ψ: The vehicle heading angle;

[0191] The control input vector u is defined as:

[0192]

[0193] v: The vehicle speed;

[0194] Δ: The front wheel steering angle;

[0195] The vehicle state update equation is based on a discrete-time dynamic model:

[0196]

[0197] where L is the wheelbase of the vehicle and Δt is the time step.

[0198] The core of the MPC controller is to optimize the control input sequence over a future period based on the current vehicle state and the reference trajectory at each control step, in order to minimize a preset cost function and satisfy the vehicle dynamics constraints.

[0199] Define the optimization objective function as a weighted sum of the state error and the control input change:

[0200]

[0201] Q is the state error weighting matrix, used to measure the deviation between the current vehicle state and the reference state:

[0202] Q = diag(q x , q y , q ψ )

[0203] q x , q y , q ψ respectively correspond to the error weights of the position x, y and the heading angle ψ.

[0204] R is the weighted matrix for controlling input changes, which is used to measure the cost of changes in control inputs (speed and steering angle):

[0205] R = diag(r v , r δ )

[0206] where r v and r δ correspond to the weights of speed and steering angle changes respectively.

[0207] Q f is the weighted matrix for terminal state error, which is used to measure the deviation between the state at the end of the prediction interval and the reference state:

[0208] Q f = diag(q f,x , q f,y , q f,ψ )

[0209] where q f,x , q f,y , q f,ψ correspond to the error weights of the predicted end coordinates x, y, and the heading angle ψ respectively.

[0210] Based on the above state update equation, ensure that the vehicle's motion conforms to the dynamic model during the optimization process. Impose speed and steering angle limits:

[0211] 0 ≤ v t ≤ v max

[0212] |δ t | ≤ δ max

[0213] v max is the maximum vehicle speed, and δ max is the maximum vehicle steering angle.

[0214] To ensure the smoothness of the acceleration process, set the upper and lower limits of the speed change rate:

[0215] |v t - v t-1 | ≤ Δv max

[0216] Δv max represents the maximum allowable value of the single-step speed change.

[0217] Based on the above objective function and constraints, construct the optimization problem:

[0218]

[0219] subject to xt+1 = A t x t + B t u t , t = 0, 1, …, T - 1

[0220] 0 ≤ v t ≤ v max

[0221] |δ t | ≤ δ max

[0222] |v t - v t-1 | ≤ Δv max , t ≥ 1

[0223] A t and B t are the linearized system matrices, x t and u t are the vehicle state vector and vehicle state respectively.

[0224] Using the cvxpy optimization library, select the solver ECOS to solve the problem and obtain the optimal control input sequence {u t}, and perform receding horizon optimization.

[0225] Based on the initial acceleration value, update the speed and position states. The speed is updated with a fixed acceleration in the first two frames:

[0226] v t = v t-1 + a initial Δt

[0227] The position is updated based on the updated speed and heading angle ψ:

[0228] x t = x t-1 + v t cos(ψ t-1 )Δt

[0229] y t = y t-1 + v t sin(ψ t-1 )Δt

[0230] Starting from the 3rd frame, based on the current vehicle state and reference trajectory, the optimal speed and steering angle for each subsequent frame are calculated sequentially by the MPC controller to update the vehicle state.

[0231] The speed component is calculated from the speed change and time interval:

[0232]

[0233] The acceleration component is calculated from the change in velocity and the time interval:

[0234]

[0235] As Figure 5 shown in the schematic diagram of the initially generated acceleration trajectory segment provided by the embodiment of the present invention, it can be seen that the simulated trajectory can well track the reference path, and the simulated trajectory shows different driving distances with different initial accelerations.

[0236] Further as an optional implementation manner, the low-speed yielding target trajectory and the initial acceleration segment target trajectory are merged to obtain a target simulated trajectory, which specifically includes:

[0237] S1051. Obtain the low-speed yielding target trajectory and the initial acceleration segment target trajectory corresponding to the same target interaction vehicle;

[0238] S1052. Determine the ending frame number of the initial acceleration segment target trajectory, and re-number the trajectory frames of the low-speed yielding target trajectory according to the ending frame number, and then add the re-numbered low-speed yielding target trajectory after the initial acceleration segment target trajectory to obtain the target simulated trajectory.

[0239] Specifically, when performing trajectory merging, sequentially traverse the sampling ID (representing the number of Monte Carlo sampling samples) in the simulator for the duration of the low-speed yielding trajectory segment of the vehicle to be turned (simulator A) and the sampling ID in the simulator for the initial acceleration of the initial acceleration segment of the vehicle to be turned (simulator B). Match and merge each pair of sampling IDs (simulator A and simulator B) one by one.

[0240] When a group of target trajectories group_B in the simulator B file is found, the target trajectory group_A in the simulator A file needs to be appended immediately after it to form continuous time-series data. To achieve this goal, the program shifts the frame number segment of group_A so that it connects to the ending frame number of group_B to form a continuous merged trajectory segment. It can be expressed as:

[0241] last_og_frame = max(groupB[og_frame]) + 1

[0242] frame_offset = last_og_frame - min(groupA[frame])

[0243] groupA[frame] ← groupA[frame] + frame_offset

[0244] Extract the trajectory information of the straight - running vehicles and all surrounding background vehicles within the time period when the vehicle to be transferred is located, and make corresponding marks.

[0245] After the merged trajectory completes the frame sequence translation and the splicing of multi - vehicle trajectory information, a complete trajectory data set for each scenario is formed. The final key fields are:

[0246] recordingId: Identify the original video recording session or number;

[0247] trackId: Vehicle number;

[0248] frame: The merged frame number;

[0249] x, y: The position of the vehicle in the plane coordinate system;

[0250] vx, vy, v: Velocity components and scalar;

[0251] psi_rad: Heading angle;

[0252] interesting_agent, track_to_predict: Vehicle role mark;

[0253] sim_a_t: Scenario merge identifier;

[0254] ms: Timestamp;

[0255] width, length: Vehicle physical dimensions.

[0256] Furthermore, as an optional implementation, determine the risk level of the target simulation trajectory, which specifically includes:

[0257] S1053. Determine the vehicle danger level of the target interaction vehicle based on the post - encroachment time;

[0258] S1054. Determine the risk perception ability of the target interaction vehicle based on the driving risk field;

[0259] Comprehensively evaluate the risk level of the target simulation trajectory according to the vehicle danger level and the risk perception ability.

[0260] Specifically, the post - encroachment time refers to the difference between the time when the first vehicle leaves the intersection point and the time when the second vehicle arrives at this point. The smaller the value, the higher the vehicle danger level:

[0261] PET = t2 - t1

[0262] t2 is the moment when the first vehicle arrives at / leaves the conflict point (or the nearest point):

[0263]

[0264] t1 is the time when the second vehicle reaches the conflict point (or the nearest point):

[0265]

[0266] Figure 6 and Figure 7 respectively show the schematic diagrams of the minimum PET value distribution and the cumulative value of PET ≤ 2s for 1000 pairs of target simulation trajectories generated by the simulator.

[0267] The driving risk field models the driver's subjective belief in risk perception using a Gaussian distribution and obtains the complete DRF spatial distribution. The main formulas are as follows:

[0268]

[0269] a: The height of the Gaussian distribution, representing the maximum value of the risk;

[0270] σ: The width of the Gaussian distribution, determining the range of risk diffusion;

[0271] x c , y c : The center of the vehicle's driving path;

[0272] R car : The turning radius of the vehicle's driving;

[0273] The distance from the point (x, y) to the path center.

[0274] Among them, the Gaussian distribution height a:

[0275] a(s) = p · (s - v · t la ) 2

[0276] p: The adjustment parameter, controlling the steepness of the risk height change;

[0277] s: The path arc length, representing the driving distance of the vehicle from the current position to the target point;

[0278] v: The speed of the vehicle;

[0279] t la : The look-ahead time, representing the driver's prediction range.

[0280] And, the width σ of the Gaussian distribution:

[0281] σ = m + k · |δ| · s + c

[0282] m: The reference width, the initial width when the vehicle is driving straight;

[0283] k: Response coefficient, representing the sensitivity of the steering angle to the width;

[0284] |δ|: Steering angle;

[0285] s: Path arc length;

[0286] C: Offset constant, related to the vehicle width.

[0287] Figure 8 The schematic diagram of the DRF distribution of an interactive scenario (including tens of thousands of target simulation trajectories) generated by the simulator is shown. PET and DRF are complementary in risk assessment. PET mainly focuses on the time safety boundary between vehicles and is suitable for capturing fine-grained timing conflicts; while DRF comprehensively considers multi-dimensional factors such as position and speed and is suitable for the quantitative assessment of overall spatial risks. The combination of the two can comprehensively evaluate the safety of the generated trajectories from multiple dimensions. By integrating the results of PET and DRF, the vehicle danger level of the target interactive vehicle is determined based on the post-encroachment time, the risk perception ability of the target interactive vehicle is determined based on the driving risk field, and the risk level of the target simulation trajectory is comprehensively evaluated according to the vehicle danger level and the risk perception ability. Based on the simulation experiment, it can be seen that the simulator finally generates both trajectories with relatively low risks and a small number of high-risk trajectories.

[0288] The method steps and experimental results of the embodiments of the present invention are described above. It can be recognized that the embodiments of the present invention extract the data of the low-speed yielding trajectory segment and the initial acceleration trajectory segment of the target interactive vehicle from the high-interaction vehicle trajectory data, respectively perform data distribution fitting to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration, and use Monte Carlo simulation to generate the target duration and the target initial acceleration according to the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration, so as to adjust the data of the low-speed yielding trajectory segment and the initial acceleration trajectory segment, obtain the low-speed yielding target trajectory and the initial acceleration segment target trajectory, and then merge them to obtain the target simulation trajectory, which can generate vehicle simulation trajectories with diverse risk levels and improve the efficiency and reliability of trajectory simulation.

[0289] Referring to Figure 9 , the embodiments of the present invention provide a diverse risk level trajectory simulation system based on Monte Carlo simulation, including:

[0290] A trajectory data extraction module, configured to obtain high-interaction vehicle trajectory data and extract the data of the low-speed yielding trajectory segment and the initial acceleration trajectory segment of the target interactive vehicle according to the high-interaction vehicle trajectory data;

[0291] A data distribution fitting module, configured to respectively perform data distribution fitting on the data of the low-speed yielding trajectory segment and the initial acceleration trajectory segment to obtain the optimal distribution parameters of the yielding time and the optimal distribution parameters of the initial acceleration;

[0292] A low-speed yielding target trajectory generation module, which is used to generate a target duration by using Monte Carlo simulation according to the low-speed yielding trajectory segment data and the optimal distribution parameters of the yielding time, and adjust the number of frames of the low-speed yielding trajectory segment data according to the target duration to obtain a low-speed yielding target trajectory;

[0293] An initial acceleration segment target trajectory generation module, which is used to generate a target initial acceleration by using Monte Carlo simulation according to the initial acceleration trajectory segment data and the optimal distribution parameters of the initial acceleration, and use an MPC controller to track a reference path according to the target initial acceleration to obtain an initial acceleration segment target trajectory;

[0294] A trajectory merging module, which is used to merge the low-speed yielding target trajectory and the initial acceleration segment target trajectory to obtain a target simulation trajectory and determine the risk level of the target simulation trajectory.

[0295] The content in the above method embodiments is applicable to the system embodiments of the present invention. The functions specifically implemented by the system embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0296] Referring to Figure 10 , an embodiment of the present invention provides a diversified risk level trajectory simulation device based on Monte Carlo simulation, including:

[0297] At least one processor;

[0298] At least one memory, which is used to store at least one program;

[0299] When the above at least one program is executed by the above at least one processor, the above at least one processor implements the above-mentioned diversified risk level trajectory simulation method based on Monte Carlo simulation.

[0300] The content in the above method embodiments is applicable to the device embodiments of the present invention. The functions specifically implemented by the device embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0301] An embodiment of the present invention also provides a computer-readable storage medium, in which a processor-executable program is stored, and the processor-executable program is used to execute the above-mentioned diversified risk level trajectory simulation method based on Monte Carlo simulation when executed by a processor.

[0302] A computer-readable storage medium according to an embodiment of the present invention can execute a diversified risk level trajectory simulation method provided by an embodiment of the method of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0303] Embodiments of the present invention also disclose a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 1 the method shown.

[0304] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order mentioned in the operation diagrams. For example, depending on the functions / operations involved, two consecutive blocks shown may actually be executed substantially simultaneously or the above-mentioned blocks can sometimes be executed in the reverse order. In addition, the embodiments presented and described in the flowcharts of the present invention are provided by way of example for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated, in which the order of various operations is changed and the sub-operations described as part of a larger operation are executed independently.

[0305] In addition, although the present invention has been described in the context of functional modules, it should be understood that unless otherwise stated to the contrary, one or more of the above functions and / or features may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It can also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More precisely, considering the attributes, functions and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the modules will be understood within the ordinary skills of an engineer. Therefore, those skilled in the art can implement the present invention as set forth in the claims without undue experimentation. It can also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0306] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the above methods in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0307] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in combination with an instruction execution system, apparatus, or device.

[0308] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part with one or more wirings (electronic device), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which the above program can be printed, because the above program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or processing it in other suitable ways as necessary, and then storing it in a computer memory.

[0309] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logic functions on data signals, application specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0310] In the above description of this specification, the description referring to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples", etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0311] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the claims and their equivalents.

[0312] The above is a specific description of the preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A diversified risk level trajectory simulation method based on Monte Carlo simulation, characterized in that: The following steps are involved: Acquire high-interaction vehicle trajectory data, and extract low-speed yielding trajectory segment data and initial acceleration trajectory segment data of the target interactive vehicle according to the high-interaction vehicle trajectory data; Performing data distribution fitting on the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data respectively to obtain optimal distribution parameters of yielding time and optimal distribution parameters of initial acceleration; Generate a target duration using Monte Carlo simulation according to the low-speed yield trajectory segment data and the yield time optimal distribution parameter, and adjust the number of frames of the low-speed yield trajectory segment data according to the target duration to obtain a low-speed yield target trajectory; Generate a target initial acceleration using Monte Carlo simulation according to the initial acceleration trajectory segment data and the initial acceleration optimal distribution parameters, and track a reference path using an MPC controller according to the target initial acceleration to obtain a target trajectory for the initial acceleration segment; The low-speed yield target trajectory and the initial acceleration segment target trajectory are combined to obtain a target simulation trajectory, and a risk level of the target simulation trajectory is determined.

2. The method for simulating diversified risk level trajectories based on Monte Carlo simulation according to claim 1, characterized in that: The step of acquiring high-interaction vehicle trajectory data and extracting low-speed yielding trajectory segment data and initial acceleration trajectory segment data of the target interactive vehicle according to the high-interaction vehicle trajectory data specifically includes: Acquire trajectory data of a target interactive vehicle that passes through an intersection and whose trajectory path is crossed by surrounding vehicles, and obtain initial trajectory data; Performing data cleaning and trajectory marking on the initial trajectory data to obtain the high-interaction vehicle trajectory data; Extracting the low-speed yielding trajectory frame segments of the target interactive vehicle whose speed is lower than a preset speed threshold according to the high-interaction vehicle trajectory data, and determining the corresponding continuous frame number and duration, to obtain the low-speed yielding trajectory segment data; An initial acceleration trajectory frame segment of a target interactive vehicle whose speed reaches the preset speed threshold for the first time is extracted according to the high-interaction vehicle trajectory data, and a corresponding initial acceleration is determined to obtain the initial acceleration trajectory segment data.

3. The diversified risk level trajectory simulation method based on Monte Carlo simulation according to claim 1 is characterized in that: The data distribution fitting is performed on the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data respectively to obtain the yielding time optimal distribution parameters and the initial acceleration optimal distribution parameters, which specifically includes: Using lognormal distribution and gamma distribution to perform data distribution fitting on the low-speed yield trajectory segment data, and determining the optimal distribution type and corresponding fitting parameters based on preset evaluation indicators to obtain the optimal distribution parameters of the yield time; Using lognormal distribution and gamma distribution to perform data distribution fitting on the initial acceleration trajectory segment data, and determining the optimal distribution type and corresponding fitting parameters based on preset evaluation indicators to obtain the optimal distribution parameters of the initial acceleration; If the optimal distribution type is lognormal distribution, the corresponding fitting parameters include log mean and log standard deviation; if the optimal distribution type is gamma distribution, the corresponding fitting parameters include shape parameter and scale parameter.

4. The method for simulating diversified risk level trajectories based on Monte Carlo simulation according to claim 2, characterized in that: The method of generating a target duration by Monte Carlo simulation according to the low-speed yield trajectory segment data and the yield time optimal distribution parameter, and adjusting the number of frames of the low-speed yield trajectory segment data according to the target duration to obtain a low-speed yield target trajectory specifically includes: According to the optimal distribution parameters of the yield time, the low-speed yield trajectory segment data is randomly sampled by using Monte Carlo simulation, and the target duration is obtained according to the sampling results. When the duration of the low-speed yield trajectory frame segment is less than the target duration, interpolating the low-speed yield trajectory frame segment to obtain an adjusted low-speed yield trajectory frame segment; When the duration of the low-speed yield trajectory frame segment is greater than the target duration, the low-speed yield trajectory frame segment is clipped to obtain an adjusted low-speed yield trajectory frame segment; The adjusted low-speed yielding trajectory frame segment is subjected to distance correction to obtain the low-speed yielding target trajectory.

5. The method for simulating diversified risk level trajectories based on Monte Carlo simulation according to claim 2, characterized in that: The method of generating a target initial acceleration by using Monte Carlo simulation according to the initial acceleration trajectory segment data and the initial acceleration optimal distribution parameters, and tracking a reference path by using an MPC controller according to the target initial acceleration to obtain a target trajectory of the initial acceleration segment specifically includes: According to the optimal distribution parameters of the initial acceleration, the initial acceleration trajectory segment data is randomly sampled by using Monte Carlo simulation, and the target initial acceleration is obtained according to the sampling results; Determining an acceleration start frame according to the initial acceleration trajectory frame segment, and determining an initial state of the vehicle corresponding to the acceleration start frame; According to the target initial acceleration and the vehicle initial state, the vehicle state of the target interactive vehicle is optimized and updated by using an MPC controller to generate a simulation trajectory frame; The low-speed yield target trajectory is obtained according to the acceleration start frame and the simulation trajectory frame.

6. The method for simulating diversified risk level trajectories based on Monte Carlo simulation according to claim 1, characterized in that: The step of merging the low-speed yield target trajectory and the initial acceleration segment target trajectory to obtain a target simulation trajectory specifically includes: Acquire the low-speed yield target trajectory and the initial acceleration segment target trajectory corresponding to the same target interactive vehicle; The end frame number of the initial acceleration segment target trajectory is determined, and the trajectory frames of the low-speed yield target trajectory are renumbered according to the end frame number, and then the renumbered low-speed yield target trajectory is added after the initial acceleration segment target trajectory to obtain the target simulation trajectory.

7. A diversified risk level trajectory simulation method based on Monte Carlo simulation according to any one of claims 1 to 6, characterized in that: Determining the risk level of the target simulation trajectory specifically includes: determining a vehicle danger level of the target interactive vehicle based on the post-encroachment time; Determining the risk perception capability of the target interactive vehicle based on the driving risk field; The risk level of the target simulation trajectory is comprehensively evaluated according to the vehicle danger level and the risk perception capability.

8. A diversified risk level trajectory simulation system based on Monte Carlo simulation, characterized in that: include: A trajectory data extraction module is used to obtain high-interaction vehicle trajectory data, and extract low-speed yielding trajectory segment data and initial acceleration trajectory segment data of the target interactive vehicle according to the high-interaction vehicle trajectory data; A data distribution fitting module is used to perform data distribution fitting on the low-speed yielding trajectory segment data and the initial acceleration trajectory segment data respectively to obtain optimal distribution parameters of yielding time and optimal distribution parameters of initial acceleration; a low-speed yield target trajectory generation module, configured to generate a target duration using Monte Carlo simulation according to the low-speed yield trajectory segment data and the yield time optimal distribution parameter, and to adjust the number of frames of the low-speed yield trajectory segment data according to the target duration to obtain a low-speed yield target trajectory; An initial acceleration segment target trajectory generation module is used to generate a target initial acceleration using Monte Carlo simulation according to the initial acceleration trajectory segment data and the initial acceleration optimal distribution parameters, and to track a reference path using an MPC controller according to the target initial acceleration to obtain an initial acceleration segment target trajectory; The trajectory merging module is used to merge the low-speed yield target trajectory and the initial acceleration segment target trajectory to obtain a target simulation trajectory, and determine the risk level of the target simulation trajectory.

9. A diversified risk level trajectory simulation device based on Monte Carlo simulation, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a diversified risk level trajectory simulation method based on Monte Carlo simulation as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to execute a diversified risk level trajectory simulation method based on Monte Carlo simulation as described in any one of claims 1 to 7 when executed by the processor.

Images