Driving style based car following behavior prediction method, device and equipment

Abstract

The present specification relates to the technical field of intelligent transportation, and provides a car-following behavior prediction method, device and equipment based on driving style, which comprises the following steps: acquiring current traffic environment parameters and trajectory data of a target vehicle; calibrating time-varying IDM parameters of a corresponding target driver according to the trajectory data; determining a driving aggressiveness index of the target driver according to the time-varying IDM parameters; determining an implicitly expressed driving style of the target driver according to the driving aggressiveness index; inputting the driving style and the current traffic environment parameters into a pre-trained car-following model based on neural processes to obtain a car-following behavior of the target driver in the current traffic environment parameters with the driving style. The embodiment of the present specification can improve the prediction accuracy of car-following behavior.

Description

Technical Field

[0001] This specification relates to the field of intelligent transportation technology, and in particular to a method, device and equipment for predicting car-following behavior based on driving style. Background Technology

[0002] In the field of intelligent transportation, car-following is the most fundamental micro-driving behavior, describing the interaction between adjacent vehicles in a convoy traveling on a one-way street with overtaking restrictions. Car-following models use dynamic methods to study the corresponding behavior of following vehicles (FVs) caused by changes in the motion state of the leading vehicle (LV). By analyzing how each vehicle follows the other, they help to understand the characteristics of single-lane traffic flow, thus bridging the gap between micro-driving behavior and macro-traffic phenomena. However, current car-following models struggle to accurately predict car-following behavior. Summary of the Invention

[0003] The purpose of the embodiments in this specification is to provide a method, apparatus, and device for predicting car-following behavior based on driving style, so as to improve the prediction accuracy of car-following behavior.

[0004] To achieve the above objectives, in one aspect, embodiments of this specification provide a method for predicting car-following behavior based on driving style, including:

[0005] Obtain the current traffic environment parameters and trajectory data of the target vehicle;

[0006] The time-varying IDM parameters of the target driver are calibrated based on the trajectory data.

[0007] The driving aggression index of the target driver is determined based on time-varying IDM parameters;

[0008] The implicit driving style of the target driver is determined based on the driving aggression index;

[0009] The driving style and the current traffic environment parameters are input into a pre-trained neural process-based car-following model to obtain the car-following behavior of the target driver under the current traffic environment parameters with the driving style.

[0010] In the driving style-based car-following behavior prediction method of this specification embodiment, the corresponding time-varying IDM parameters are calibrated according to the trajectory data, including:

[0011] Step a: Select a data frame to be processed from the trajectory data as the current frame;

[0012] Step b: Determine the prior distribution of the current frame;

[0013] Step c: Sample a set of time-varying parameters from the prior distribution of the current frame;

[0014] Step d: Input the set of time-varying parameters into the IDM model to calculate the acceleration and desired headway of the target vehicle;

[0015] Step e: Determine whether the acceleration and the desired headway meet the preset conditions;

[0016] Step f: When the acceleration and the desired headway meet the preset conditions, save the set of time-varying parameters;

[0017] Step g: Repeat steps b to f above until the number of time-varying parameter groups of the current frame that are currently saved reaches the set threshold.

[0018] Step h: Fit the posterior distribution of the current frame based on the multiple time-varying parameters of the current frame, and use the posterior distribution as the time-varying IDM parameter of the current frame;

[0019] Step i: Repeat steps a to h above until the time-varying IDM parameters of each frame in the trajectory data are obtained.

[0020] In the driving style-based car-following behavior prediction method of this specification embodiment, determining the prior distribution of the current frame includes:

[0021] If the current frame is the first frame in the trajectory data, define the prior distribution of the first frame;

[0022] If the current frame is not the first frame in the trajectory data, the posterior distribution of the previous frame is used as the prior distribution of the current frame.

[0023] In the driving style-based car-following behavior prediction method of this specification embodiment, the preset conditions include:

[0024] The error between the acceleration and the acceleration in the current frame is less than a first error threshold; and the error between the expected front-end spacing and the front-end spacing in the current frame is less than a second error threshold.

[0025] In the driving style-based car-following behavior prediction method of the embodiments of this specification, the IDM model includes:

[0026]

[0027] Among them, a n (t) represents the acceleration of the target vehicle at time t, a max,n (t) represents the desired acceleration of the target vehicle at time t, v n (t) represents the velocity of the target vehicle at time t, v 0,n(t) represents the desired speed of the target vehicle at time t, δ is the hyperparameter, and Δv n (t) represents the speed difference between the target vehicle and the vehicle in front at time t, s * (v n (t),Δv n (t) represents the expected headway of the target vehicle at time t, and s n (t) represents the headway of the target vehicle at time t, s 0,n (t) represents the headway of the target vehicle when it stops at time t, and max is the maximum value function. n (t) represents the expected headway of the target vehicle at time t, b n (t) represents the desired deceleration of the target vehicle at time t.

[0028] In the driving style-based car-following behavior prediction method of this specification embodiment, determining the target driver's driving aggression index based on time-varying IDM parameters includes:

[0029] The driving aggression index of the target driver is calculated by inputting the time-varying IDM parameters into the following formula.

[0030]

[0031] Where H represents the aggressive driving index, and R... i Let M(θ) be the influence weight of the i-th time-varying IDM parameter, where θ is the time-varying IDM parameter. i Let Q1(θ) be the mean of the i-th time-varying IDM parameter. i Let Q3(θ) be the quarter quantile of the i-th time-varying IDM parameter. i For the i-th time-varying IDM parameter, M(θ') is a three-quarters position. + ) i Let S(θ') be the mean of the positive-zero portion of the time-series difference sequence of the i-th time-varying IDM parameter. + ) i M(θ') represents the standard deviation of the positive-zero portion of the time-series difference sequence of the i-th time-varying IDM parameter. - ) i S(θ') is the mean of the portion of the time-series difference sequence of the i-th time-varying IDM parameter that is less than or equal to zero. - ) i Let be the standard deviation of the portion of the time-series difference sequence of the i-th time-varying IDM parameter that is less than or equal to zero, V0 be the expected speed of the target vehicle, T be the expected headway of the target vehicle, S0 be the expected headway of the target vehicle, and a be the expected headway of the target vehicle. max Let b be the desired acceleration of the target vehicle, and let b be the desired deceleration of the target vehicle.

[0032] In the driving style-based car-following behavior prediction method of this specification embodiment, determining the implicitly expressed driving style of the target driver based on the driving aggression index includes:

[0033] Input the driving aggression index into the formula The dimensionality-reduced implicit expression of the target driver's driving style is calculated; where, H represents the implicit expression of the target driver's driving style after dimensionality reduction, where H is the driving aggression index, α is the coefficient, and β is a constant.

[0034] The driving style of the target driver is reconstructed from the dimensionality-reduced implicit expression of the target driver using the inverse principal component analysis algorithm.

[0035] On the other hand, embodiments of this specification also provide a following behavior prediction device based on driving style, including:

[0036] The basic parameter acquisition module is used to acquire the current traffic environment parameters and trajectory data of the target vehicle.

[0037] The IDM parameter calibration module is used to calibrate the time-varying IDM parameters of the target driver based on the trajectory data.

[0038] The aggressiveness index determination module is used to determine the driving aggressiveness index of the target driver based on the time-varying IDM parameters;

[0039] A driving style determination module is used to determine the implicit driving style of the target driver based on the driving aggression index.

[0040] The car-following behavior prediction module is used to input the driving style and the current traffic environment parameters into a pre-trained neural process-based car-following model to obtain the car-following behavior of the target driver under the current traffic environment parameters with the driving style.

[0041] On the other hand, embodiments of this specification also provide a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the computer program, when run by the processor, executes instructions for the above-described method.

[0042] On the other hand, embodiments of this specification also provide a computer storage medium storing a computer program thereon, which, when run by the processor of a computer device, executes instructions for the above-described method.

[0043] As can be seen from the technical solutions provided in the embodiments of this specification above, the embodiments of this specification pre-design an IDM model with time-varying parameters, train a car-following model based on neural processes, and construct a mapping relationship between the time-varying IDM parameters and the intermediate latent variables (i.e., implicitly expressed driving styles) of the NP model. During prediction, when driving environment parameters and trajectory data are obtained, the corresponding time-varying IDM parameters can be calibrated from the IDM model. Since these time-varying IDM parameters can express both heterogeneity among drivers (i.e., different driving styles of different drivers) and heterogeneity within drivers (i.e., changes in driving style of the same driver), they can accurately capture the driver's driving style. Based on this, the intermediate latent variables of the NP model can be obtained according to the above mapping relationship. By inputting these intermediate latent variables and driving environment parameters into the NP model, accurate prediction of car-following behavior under any given driving style can be achieved. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0045] Figures 1a-1d This specification shows schematic diagrams illustrating application scenarios of the car-following model in some embodiments;

[0046] Figure 2 Flowcharts of driving style-based car-following behavior prediction methods in some embodiments of this specification are shown;

[0047] Figure 3 This document illustrates a flowchart of a driving style-based car-following behavior prediction method for some embodiments of this specification, in which time-varying IDM parameters are calibrated based on trajectory data.

[0048] Figure 4 This specification shows a schematic diagram of a car-following model training based on neural processes in an exemplary embodiment.

[0049] Figure 5 This diagram illustrates a prediction using a pre-trained car-following model based on neural processes in an exemplary embodiment of this specification.

[0050] Figure 6 This specification shows a structural block diagram of a car-following behavior prediction device based on driving style in some embodiments;

[0051] Figure 7A structural block diagram of a computer device in some embodiments of this specification is shown.

[0052] [Explanation of Labels in the Attached Image]

[0053] 10. Automated driving system;

[0054] 20. Autonomous driving simulation system;

[0055] 30. Intelligent Transportation Systems;

[0056] 40. Intelligent Transportation Simulation System;

[0057] 61. Basic Parameter Acquisition Module;

[0058] 62. IDM parameter calibration module;

[0059] 63. Radical Index Determination Module;

[0060] 64. Driving style determination module;

[0061] 65. Car-following behavior prediction module;

[0062] 702. Computer equipment;

[0063] 704, Processor;

[0064] 706. Memory;

[0065] 708. Drive mechanism;

[0066] 710. Input / output interfaces;

[0067] 712. Input devices;

[0068] 714. Output devices;

[0069] 716. Presentation equipment;

[0070] 718. Graphical User Interface;

[0071] 720. Network interface;

[0072] 722. Communication link;

[0073] 724. Communication bus. Detailed Implementation

[0074] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0075] Traditional car-following models are mostly based on standard driving behavior and fail to accurately reflect individual vehicle differences. Therefore, they struggle to accurately predict car-following behavior. For example, the traditional Intelligent Driver Model (IDM) is a longitudinal traffic flow model containing only a few parameters with clear physical meaning. Traditional IDM models can describe the different state changes between free flow and congested flow using a uniform model form. However, traditional IDM models lack a stochastic component; that is, the output is deterministic when the input is fixed, which differs from the randomness of real-world vehicle behavior. For instance, in traffic flow simulations, two vehicles with identical parameters can be observed starting simultaneously at the stop line at an intersection and traveling parallel to each other, maintaining the same driving state for a considerable period. This does not match actual vehicle driving behavior. In some embodiments of this specification, car-following behavior can be understood as the acceleration determined by the vehicle given a specific driving style and traffic environment.

[0076] In view of this, in order to improve the prediction accuracy of car-following behavior, this specification provides a new car-following model and a technical solution for predicting car-following behavior based on this new car-following model. In this specification, an IDM model with time-varying parameters is pre-designed, a car-following model based on Neural Process (NP) (hereinafter referred to as the NP model) is trained, and a mapping relationship is constructed between the time-varying IDM parameters and the intermediate latent variables (i.e., implicitly expressed driving styles) of the NP model. During prediction, when driving environment parameters and trajectory data are obtained, the corresponding time-varying IDM parameters can be calibrated from the IDM model. Since these time-varying IDM parameters (i.e., IDM parameters that change over time) can express both heterogeneity among drivers (i.e., different driving styles of different drivers) and heterogeneity within drivers (i.e., changes in the driving style of the same driver), the driver's driving style can be accurately captured. Based on this, the intermediate latent variables of the NP model can be obtained according to the above mapping relationship. By inputting these intermediate latent variables and driving environment parameters into the NP model, accurate prediction of car-following behavior under any given driving style can be achieved.

[0077] like Figures 1a-1dAs shown, the car-following model and the technical solution for predicting car-following behavior based on the new car-following model in the embodiments of this specification can be applied to autonomous driving system 10, autonomous driving simulation system 20, intelligent transportation system 30, intelligent transportation simulation system 40, etc.

[0078] This specification provides an embodiment of a car-following behavior prediction method based on driving style, which can be applied to the aforementioned autonomous driving systems, autonomous driving simulation systems, intelligent transportation systems, and intelligent transportation simulation systems. (Reference) Figure 2 As shown, in some embodiments, the car-following behavior prediction method based on driving style may include the following steps:

[0079] Step 201: Obtain the current traffic environment parameters and trajectory data of the target vehicle.

[0080] The target vehicle is the vehicle currently selected for processing. In a real driving environment, the current traffic environment parameters and trajectory data of the target vehicle can be obtained through various related devices installed on the vehicle (such as satellite positioning and navigation systems, accelerometers, speed sensors, lidar, visual radar, etc.). In a simulated driving environment, the current traffic environment parameters and trajectory data of the target vehicle can be obtained from the corresponding simulation system.

[0081] Current traffic environment parameters are used to quantitatively describe the traffic environment situation of the target vehicle at the current moment. In some embodiments, traffic environment parameters may include Δv. n (t), v n (t), s n (t), v n-1 (t), Where, Δv n (t) represents the speed difference between the target vehicle and the vehicle in front at time t, v n (t) represents the velocity of the target vehicle at time t, v n-1 (t) is the speed of the vehicle in front at time t. and These are the lateral and longitudinal accelerations of the vehicle in front at time t, respectively.

[0082] Trajectory data refers to the movement trajectory data of a target vehicle from its starting point to the present during a one-way driving process. In some embodiments, trajectory data may include sampling time, vehicle ID, vehicle position (e.g., vehicle latitude and longitude), vehicle position speed, etc.

[0083] Step 202: Calibrate the time-varying IDM parameters of the corresponding target driver based on the trajectory data.

[0084] The target driver refers to the driver currently driving the target vehicle. The IDM model with time-varying parameters in this embodiment is defined as follows:

[0085]

[0086] Among them, a n (t) represents the acceleration of the target vehicle at time t, a max，n (t) represents the desired acceleration of the target vehicle at time t, v n (T) represents the velocity of the target vehicle at time t, v 0，n (T) represents the desired speed of the target vehicle at time t, δ is the hyperparameter, and Δv n (t) represents the speed difference between the target vehicle and the vehicle in front at time t, s * (v n (T), Δv n (T) represents the expected headway of the target vehicle at time t, and s n (t) represents the headway of the target vehicle at time t, s 0，n (t) represents the headway of the target vehicle when it stops at time T, and max is the maximum value function, T n (t) represents the expected headway of the target vehicle at time t, b n (t) represents the desired deceleration of the target vehicle at time t.

[0087] In the aforementioned IDM model with time-varying parameters, the time-varying parameters that need to be calibrated include v. 0,n (t), T n (t), s 0,n (t), a max,n (t) and b n (t); the independent variables of the model include Δv n (t) and Δv n (t), the model output is a n (t).

[0088] In the embodiments of this specification, the time-varying IDM parameters obtained by calibrating using the trajectory data of the target driver can be used to represent the driver's driving style; for example, a larger expected speed and a smaller expected headway usually indicate that the driver's driving style is more aggressive; a smaller expected speed and a larger expected headway usually indicate that the driver's driving style is more cautious.

[0089] Because different drivers have different driving styles, and the driving style of the same driver may also change at different times or in different scenarios, driving style has a randomness. Therefore, to characterize the randomness of driving style, each time-varying IDM parameter can be modeled as a stochastic process.

[0090] Current calibration algorithms are not suitable for calibrating time-varying IDM parameters. Therefore, a new calibration algorithm is designed in the embodiments of this specification. It is based on Approximate Bayesian Computation (ABC). In this new calibration algorithm, it is assumed that each time-varying IDM parameter is independent, and multi-model fitting is used to determine the model of the prior distribution. The initial prior distribution of each time-varying IDM parameter is defined as a Gaussian distribution, with the expected value being a fixed parameter obtained by calibration using traditional calibration methods and a set of data, and the variance being a predefined hyperparameter. Assuming that the IDM parameter at time t is affected by the IDM parameter at time t-1, when calibrating the IDM parameter at time t, we use the posterior distribution at time t-1 as the prior distribution for calibrating the IDM parameter at time t. The specific steps for calibrating the time-varying IDM parameters of the corresponding target driver using the new calibration algorithm and trajectory data will be described below.

[0091] Step 203: Determine the driving aggression index of the target driver based on the time-varying IDM parameters.

[0092] To accurately describe a driver's driving style, this specification proposes the concept of a driving aggression index (or aggression index) in its embodiments. The driving aggression index can be used to quantitatively describe (or explain) the degree of a driver's aggressiveness in driving a vehicle. Specifically, a mapping relationship between time-varying IDM parameters and the driving aggression index can be pre-established, and then the calibrated time-varying IDM parameters can be input into the mapping relationship to obtain the corresponding driving aggression index.

[0093] In some embodiments, determining the target driver's driving aggression index based on time-varying IDM parameters may include:

[0094] The driving aggression index of the target driver is calculated by inputting the time-varying IDM parameters into the following formula.

[0095]

[0096] Where H represents the aggressive driving index, and R... i Let M(θ) be the influence weight of the i-th time-varying IDM parameter, where θ is the time-varying IDM parameter. i Let Q1(θ) be the mean of the i-th time-varying IDM parameter. i Let Q3(θ) be the quantile of the i-th time-varying IDM parameter. i For the i-th time-varying IDM parameter, M(θ') is a three-quarters position. + ) iLet S(θ') be the mean of the positive-zero portion of the time-series difference sequence of the i-th time-varying IDM parameter. + ) i M(θ') represents the standard deviation of the positive-zero portion of the time-series difference sequence of the i-th time-varying IDM parameter. - ) i S(θ') is the mean of the portion of the time-series difference sequence of the i-th time-varying IDM parameter that is less than or equal to zero. - ) i Let be the standard deviation of the portion of the time-series difference sequence of the i-th time-varying IDM parameter that is less than or equal to zero, V0 be the expected speed of the target vehicle, T be the expected headway of the target vehicle, S0 be the expected headway of the target vehicle, and a be the expected headway of the target vehicle. max Let θ be the desired acceleration of the target vehicle, and b be the desired deceleration of the target vehicle. θ = [V0, T, S0, a] max [,b] indicates that the time-varying IDM parameters from the 1st to the 5th are V0,T,S0,a max ,b,R i =[+1,-1,-1,+1,-1] represents V0,T,S0,a max The influence weights of b are +1, -1, -1, +1, -1.

[0097] Among them, V0 and a max Directly proportional to H, T, S0, and b are inversely proportional to H. A larger value of H indicates a more aggressive driving style. The quantiles mentioned above, also called quartiles, are numerical points that divide the probability distribution range of a random variable into several equal parts. Commonly used quantiles include the median (i.e., bisector), quartiles, and percentiles. The mean of a time-varying IDM parameter refers to the mean of all corresponding time-varying IDM parameters within the trajectory data. For example, Δv... n Taking (t) as an example, M(Δv) n (t) represents the Δv at all sampling times within the trajectory data. n The mean of (t).

[0098] Step 204: Determine the implicit driving style of the target driver based on the driving aggression index.

[0099] Since one of the inputs to the NP model is the implicitly expressed driving style, after determining the target driver's implicitly expressed driving style based on the driving aggression index, it can be input into the pre-trained NP model to predict the corresponding car-following behavior.

[0100] In some embodiments, determining the implicitly expressed driving style of the target driver based on the driving aggression index may include: first inputting the driving aggression index into a formula. The dimensionality-reduced implicit expression of the target driver's driving style is calculated; where, Let H be the dimensionality-reduced implicit expression of the target driver's driving style, where H is the driving aggression index, α is the coefficient, and β is a constant. Then, the Inverse Principal Component Analysis (PCA) algorithm is used to reconstruct the target driver's implicit driving style from the dimensionality-reduced implicit expression of the target driver's driving style.

[0101] Since the intermediate latent variables of the NP model are an implicit expression of driving style, which is difficult for drivers to understand and interpret, while time-varying IDM parameters can not only accurately capture the driver's driving style but are also easy for the driver to understand and interpret, it is necessary to construct a mapping relationship between the time-varying IDM parameters and the intermediate latent variables of the NP model (i.e., the implicitly expressed driving style) in order to predict car-following behavior under any given driving style. In some embodiments, the above-mentioned formulas for calculating the driving aggression index and the implicitly expressed driving style form the mapping relationship between the time-varying IDM parameters and the intermediate latent variables of the NP model.

[0102] Step 205: Input the driving style and the current traffic environment parameters into the pre-trained neural process-based car-following model to obtain the car-following behavior of the target driver under the current traffic environment parameters with the driving style.

[0103] The driving style of the target driver, characterized by time-varying IDM parameters represented by a set of stochastic processes, can be used to explain why the target driver makes such car-following behavior. However, since the time-varying IDM parameters require calibration of real data (i.e., trajectory data), this time-varying IDM model can only describe the target driver's driving style and cannot directly predict the target driver's car-following behavior. To address this issue, the embodiments in this specification employ a pre-trained NP model to predict the target driver's car-following behavior.

[0104] Neural processes possess a powerful ability to fit functions. Unlike other deep learning models used for function fitting, neural processes use neural networks to represent the distribution of a function rather than a single fixed function. A neural process can be modeled as follows:

[0105]

[0106] Where x i and y iThe observed data used to fit the relationship is z, which is defined as a high-dimensional latent variable that determines the functional relationship, and p represents an abstract probability distribution. This randomness inherent in neural processes is well-suited for fitting car-following behavior; that is, the neural process can be pre-trained, and the decoder of the trained neural process can be used as a car-following model to predict the driver's car-following behavior. The training process of the NP model is described below.

[0107] The car-following behavior of each driver (trajectory data from multiple drivers was used during training) is defined as a function distribution. To train the neural process to simulate car-following behavior, the input x of the neural process is defined as: Δv n (t), v n (t), s n (t), v n-1 (t), As explained above, this set of inputs x can be used to describe the current traffic environment. To train this neural process, each driver's trajectory data can be divided into context points and target points. The context points are input into the encoder to obtain their representation in the latent space. This output is further input into the aggregator to obtain latent variables. The latent variables and the input x from the target point are input into the decoder to predict the y corresponding to x (i.e., car-following behavior).

[0108] Combination Figure 4 As shown, to introduce randomness, two encoders are designed in the neural process: a deterministic encoder and a latent encoder. The deterministic encoder is used to output the context point (x). c ,y c Implicit representation of r in implicit space c (i.e. r) c (These are latent variables). Where x c This refers to the current traffic environment, y c That is, x c The driving style is r c The following behavior. Latent variable r c Because it integrates information from context points, it can be considered an implicit expression of driving style. The implicit encoder outputs: context point (x... c ,y c The probability distribution of the latent variables of ) has a mean and variance of μ. c ,σ c (μ will be used below) c ,σ c (representing the distribution), and the target point (x) T,y T The probability distribution of the latent variables of ) has a mean and variance of μ. T ,σ T (N(μ) is used below) T ,σ T (This represents the distribution). Where x represents the target point. T This refers to the current traffic environment, y T That is, x T The driving style is r c The following behavior. During training, the optimization objective of the neural process is to reduce the above two distributions (i.e., N(μ)). c ,σ c ) and N(μ T ,σ T The KL divergence between the distribution and the normal distribution (i.e., obtaining N(μ)) c ,σ c ) and N(μ T ,σ T (Approaching or tending towards a normal distribution), while increasing the target point (x) T ,y T ) in N(μ c ,σ c The probability on the distribution (i.e., increasing) Figure 4 p(y) T |x c ,y c When the above optimization objective is achieved, the decoder of the neural process is the trained car-following model based on the neural process. Figure 4 middle, Indicates output The probability distribution; express With y T The Euclidean distance is given by |||2, which represents the second norm.

[0109] Combination Figure 5 As shown, in some embodiments, the prediction is performed using driving style r and current traffic environment parameter x. * Input to a pre-trained neural process-based car-following model (i.e. Figure 5 In the decoder, the target driver's driving style r is obtained in the current traffic environment parameter x. * Following behavior exist Figure 5 middle, Indicates output The probability distribution is given by Z, where Z is the output of the implicit encoder.

[0110] refer to Figure 3 As shown, calibrating the corresponding time-varying IDM parameters based on the trajectory data may include the following steps:

[0111] Step 301: Select a data frame to be processed from the trajectory data as the current frame.

[0112] Step 302: Determine the prior distribution of the current frame.

[0113] When determining the prior distribution of the current frame, if the current frame is the first frame in the trajectory data, the prior distribution of the first frame can be defined; for example, the prior distribution of the first frame in the trajectory data can be defined as a Gaussian distribution with a preset mean and a preset variance.

[0114] If the current frame is not the first frame in the trajectory data, the posterior distribution of the previous frame is used as the prior distribution of the current frame.

[0115] Step 303: Sample a set of time-varying parameters from the prior distribution of the current frame.

[0116] Step 304: Input the set of time-varying parameters into the IDM model to calculate the acceleration of the target vehicle and the desired headway.

[0117] Step 305: Determine whether the acceleration and the desired headway meet the preset conditions; if the preset conditions are not met, proceed to step 306; if the preset conditions are met, proceed to step 307.

[0118] In some embodiments, the preset conditions may be: the error between the calculated acceleration and the acceleration in the current frame is less than a first error threshold; and the error between the calculated expected vehicle headway and the vehicle headway in the current frame is less than a second error threshold. The first and second error thresholds can be preset as needed.

[0119] Step 306: Discard the set of time-varying parameters and proceed to step 303.

[0120] Step 307: Save the set of time-varying parameters.

[0121] Step 308: Determine whether the number of time-varying parameter groups of the current frame has reached the set threshold; if the set threshold has not been reached, proceed to step 303; if the set threshold has been reached, proceed to step 309.

[0122] Step 309: Fit the posterior distribution of the current frame based on the multiple time-varying parameters of the current frame, and use the posterior distribution as the time-varying IDM parameter of the current frame.

[0123] Step 310: Determine whether the time-varying IDM parameters of all frames in the trajectory data have been obtained; if the time-varying IDM parameters of all frames in the trajectory data have not been obtained, proceed to step 301; if the time-varying IDM parameters of all frames in the trajectory data have been obtained, end the calibration.

[0124] Although the process described above includes multiple operations that occur in a specific order, it should be clearly understood that these processes may include more or fewer operations, which may be executed sequentially or in parallel (e.g., using parallel processors or a multithreaded environment).

[0125] Corresponding to the above-described method for predicting car-following behavior based on driving style, this specification also provides a device for predicting car-following behavior based on driving style, see reference. Figure 6 As shown, in some embodiments, the car-following behavior prediction device based on driving style may include:

[0126] The basic parameter acquisition module 61 can be used to acquire the current traffic environment parameters and trajectory data of the target vehicle.

[0127] The IDM parameter calibration module 62 can be used to calibrate the time-varying IDM parameters of the target driver based on the trajectory data.

[0128] The aggressiveness index determination module 63 can be used to determine the driving aggressiveness index of the target driver based on the time-varying IDM parameters;

[0129] The driving style determination module 64 can be used to determine the implicit driving style of the target driver based on the driving aggression index.

[0130] The car-following behavior prediction module 65 can be used to input the driving style and the current traffic environment parameters into a pre-trained neural process-based car-following model to obtain the car-following behavior of the target driver under the current traffic environment parameters with the driving style.

[0131] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this specification are all information and data authorized and agreed upon by the user and fully authorized by all parties.

[0133] Embodiments of this specification also provide a computer device. For example... Figure 7As shown, in some embodiments of this specification, the computer device 702 may include one or more processors 704, such as one or more central processing units (CPUs) or graphics processing units (GPUs), each of which may implement one or more hardware threads. The computer device 702 may also include any memory 706 for storing any kind of information such as code, settings, data, etc. In one specific embodiment, a computer program is stored on the memory 706 and can run on the processor 704. When the computer program is run by the processor 704, it can execute instructions of the driving style-based following behavior prediction method described in any of the above embodiments. Without limitation, for example, the memory 706 may include any type of RAM, any type of ROM, flash memory device, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory may represent a fixed or removable component of the computer device 702. In one case, when the processor 704 executes associated instructions stored in any memory or combination of memories, the computer device 702 can perform any operation of the associated instructions. The computer device 702 also includes one or more drive mechanisms 708 for interacting with any memory, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.

[0134] Computer device 702 may also include an input / output interface 710 (I / O) for receiving various inputs (via input device 712) and providing various outputs (via output device 714). A specific output mechanism may include a presentation device 716 and an associated graphical user interface 718 (GUI). In other embodiments, the input / output interface 710 (I / O), input device 712, and output device 714 may be omitted, and the device may function solely as a computer device within a network. Computer device 702 may also include one or more network interfaces 720 for exchanging data with other devices via one or more communication links 722. One or more communication buses 724 couple the components described above together.

[0135] Communication link 722 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 722 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0136] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), computer-readable storage media, and computer program products according to some embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processor to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processor, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0137] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processor to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0138] These computer program instructions may also be loaded onto a computer or other programmable data processor to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable device, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0139] In a typical configuration, a computer device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0140] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0141] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by computer equipment. As defined in this specification, computer-readable media does not include transient media, such as modulated data signals and carrier waves.

[0142] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of computer program products implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0143] The embodiments described in this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. The embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processors connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0144] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0145] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0146] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the embodiments of this specification. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0147] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for predicting car-following behavior based on driving style, characterized in that, include: Obtain the current traffic environment parameters and trajectory data of the target vehicle; The time-varying intelligent driver model (IDM) parameters of the target driver are calibrated based on the trajectory data. The driving aggression index of the target driver is determined based on time-varying IDM parameters; The implicit driving style of the target driver is determined based on the driving aggression index; The driving style and the current traffic environment parameters are input into a pre-trained neural process-based car-following model to obtain the car-following behavior of the target driver under the current traffic environment parameters with the driving style. The step of calibrating the corresponding time-varying IDM parameters based on the trajectory data includes: Step a: Select a data frame to be processed from the trajectory data as the current frame; Step b: Determine the prior distribution of the current frame, specifically including: if the current frame is the first frame in the trajectory data, define the prior distribution of the first frame; if the current frame is not the first frame in the trajectory data, use the posterior distribution of the previous frame as the prior distribution of the current frame. Step c: Sample a set of time-varying parameters from the prior distribution of the current frame; Step d: Input the set of time-varying parameters into the IDM model to calculate the acceleration and desired headway of the target vehicle; Step e: Determine whether the acceleration and the desired headway meet the preset conditions; Step f: When the acceleration and the desired headway meet the preset conditions, save the set of time-varying parameters; Step g: Repeat steps b to f above until the number of time-varying parameter groups of the current frame that are currently saved reaches the set threshold. Step h: Fit the posterior distribution of the current frame based on the multiple time-varying parameters of the current frame, and use the posterior distribution as the time-varying IDM parameter of the current frame; Step i: Repeat steps a to h above until the time-varying IDM parameters of each frame in the trajectory data are obtained.

2. The method for predicting car-following behavior based on driving style as described in claim 1, characterized in that, The preset conditions include: The error between the acceleration and the acceleration in the current frame is less than a first error threshold; and the error between the expected front-end spacing and the front-end spacing in the current frame is less than a second error threshold.

3. The method for predicting car-following behavior based on driving style as described in claim 1, characterized in that, The IDM model includes: in, For the target vehicle in acceleration at any moment For the target vehicle in Expected acceleration at any moment For the target vehicle in The speed of time For the target vehicle in Expected speed at any moment For hyperparameters, For the target vehicle in The speed difference between the current vehicle and the vehicle in front. For the target vehicle in Expected headway at any given moment For the target vehicle in The distance between the front of the car at any given time, For the target vehicle in The distance between the front of the car when it is stationary. It is a function with maximum value. For the target vehicle in Expected headway at any given moment For the target vehicle in Expected deceleration at any given moment.

4. The method for predicting car-following behavior based on driving style as described in claim 1, characterized in that, The step of determining the target driver's driving aggression index based on time-varying IDM parameters includes: The driving aggression index of the target driver is calculated by inputting the time-varying IDM parameters into the following formula. in, For aggressive driving index, For the first The influence weight of each time-varying IDM parameter For time-varying IDM parameters, For the first The mean of each time-varying IDM parameter, For the first One-quarter quantile of time-varying IDM parameters For the first Three-quarters of the time-varying IDM parameters For the first The mean of the positive-zero portions of the time-series difference sequence of each time-varying IDM parameter. For the first The standard deviation of the positive-zero portion of the time-varying IDM parameter's time-series difference sequence. For the first The mean of the less than or equal to zero portions of the time-series difference sequence of each time-varying IDM parameter. For the first The standard deviation of the portion of the time-series difference sequence of each time-varying IDM parameter that is less than or equal to zero. The desired speed of the target vehicle, The desired headway of the target vehicle. The desired front-end spacing of the target vehicle. For the target vehicle's desired acceleration, The desired deceleration for the target vehicle.

5. The method for predicting car-following behavior based on driving style as described in claim 1, characterized in that, The step of determining the implicitly expressed driving style of the target driver based on the driving aggression index includes: Input the driving aggression index into the formula The dimensionality-reduced implicit expression of the target driver's driving style is calculated; where, The implicit expression of driving style for the target driver after dimensional reduction. For aggressive driving index, coefficient It is a constant; The driving style of the target driver is reconstructed from the dimensionality-reduced implicit expression of the target driver using the inverse principal component analysis algorithm.

6. A following behavior prediction device based on driving style, characterized in that, include: The basic parameter acquisition module is used to acquire the current traffic environment parameters and trajectory data of the target vehicle. The IDM parameter calibration module is used to calibrate the time-varying intelligent driver model (IDM) parameters of the target driver based on the trajectory data. The aggressiveness index determination module is used to determine the driving aggressiveness index of the target driver based on the time-varying IDM parameters; A driving style determination module is used to determine the implicit driving style of the target driver based on the driving aggression index. The car-following behavior prediction module is used to input the driving style and the current traffic environment parameters into a pre-trained car-following model based on neural processes to obtain the car-following behavior of the target driver under the current traffic environment parameters with the driving style. The step of calibrating the corresponding time-varying IDM parameters based on the trajectory data includes: Step a: Select a data frame to be processed from the trajectory data as the current frame; Step b: Determine the prior distribution of the current frame, specifically including: if the current frame is the first frame in the trajectory data, define the prior distribution of the first frame; if the current frame is not the first frame in the trajectory data, use the posterior distribution of the previous frame as the prior distribution of the current frame. Step c: Sample a set of time-varying parameters from the prior distribution of the current frame; Step d: Input the set of time-varying parameters into the IDM model to calculate the acceleration and desired headway of the target vehicle; Step e: Determine whether the acceleration and the desired headway meet the preset conditions; Step f: When the acceleration and the desired headway meet the preset conditions, save the set of time-varying parameters; Step g: Repeat steps b to f above until the number of time-varying parameter groups of the current frame that are currently saved reaches the set threshold. Step h: Fit the posterior distribution of the current frame based on the multiple time-varying parameters of the current frame, and use the posterior distribution as the time-varying IDM parameter of the current frame; Step i: Repeat steps a to h above until the time-varying IDM parameters of each frame in the trajectory data are obtained.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, When the computer program is run by the processor, it executes the instructions of the method according to any one of claims 1-5.

8. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor of the computer device, it executes the instructions of the method according to any one of claims 1-5.

Images