A driving behavior analysis method, system and application for a highland environment

Abstract

The application provides a driving behavior analysis method and system for a plateau environment and an application, and comprises the following steps: constructing a plain and a constant-elevation plateau environment, obtaining a total mileage of plain driving, plain driving data, total mileage of plateau driving and plateau driving data; preprocessing the total mileages of plain driving and plateau driving; obtaining first and second effective driving data based on a driving performance decline time window; respectively performing probability density fitting on the first and second effective driving data; and obtaining a JS divergence value between the probability density distributions of the first and second effective driving data based on JS divergence. The application intuitively quantifies the driving difference degree between the plateau and plain environments through the JS divergence value from the perspective of driving data under constant elevation, avoids the interference of objective factors, has high accuracy in analyzing the plateau environment driving behavior, can reasonably match an auxiliary driving strategy based on the difference degree, and improves the driving safety in the plateau environment.

Description

Technical Field

[0001] This invention relates to the field of transportation technology in high-altitude environments, and more specifically, to a method, system, and application for analyzing driving behavior in high-altitude environments. Background Technology

[0002] The driving environment for drivers includes both plains and plateau environments. Driving is most efficient and comfortable at an altitude of 0m. As the altitude increases, the absolute amount of oxygen gradually decreases, and the blood oxygen concentration in the human body decreases, which has a negative impact on the driver's condition. Drivers are prone to frequent yawning and fatigue. Driving at high altitudes requires more caution and involves a greater driving load. Driving behavior in high-altitude environments is greatly affected. Analyzing driving behavior in high-altitude environments is of great significance for improving driving safety in high-altitude environments.

[0003] Existing technologies typically explore the impact of high-altitude environments on drivers' states from a biopsychological perspective, such as electroencephalography (EEG), electrocardiography (ECG), electromyography (EMG), and fatigue levels, through real-vehicle tests. They indirectly analyze driving behavior by analyzing the changes in EEG, ECG, EMG, and fatigue levels. However, this approach cannot provide a direct analysis of driving behavior in high-altitude environments from a data perspective. Furthermore, real-vehicle tests are easily affected by objective factors, resulting in low data accuracy and making it impossible to accurately analyze driving behavior in high-altitude environments. Moreover, real-vehicle tests involve changes in altitude, making it impossible to analyze driving behavior at a constant altitude in high-altitude environments. Summary of the Invention

[0004] This invention provides a method, system, and application for analyzing driving behavior in high-altitude environments. Under constant altitude conditions, it provides an intuitive analysis of driving behavior in high-altitude environments from the perspective of driving data, avoiding the problem of reduced accuracy of driving behavior analysis in high-altitude environments due to interference from objective factors when using real vehicle testing.

[0005] On one hand, this invention provides a driving behavior analysis method for high-altitude environments, comprising the following steps: S1, based on simulated real-vehicle driving tests, constructing plain environments and constant-altitude high-altitude environments respectively, and obtaining the total plain driving mileage, plain driving data, total high-altitude driving mileage, and high-altitude driving data under preset driving time; S2, preprocessing the total plain driving mileage and total high-altitude driving mileage to obtain preprocessed plain driving mileage and preprocessed high-altitude driving mileage; S3, based on the time window of driving performance decline in high-altitude environments, obtaining the effective mileage interval, and recording the driving data collected from the preprocessed high-altitude driving mileage within the effective mileage interval as... The first effective driving data is the driving data collected from the preprocessed plain driving mileage within the effective mileage range, which is recorded as the second effective driving data; S4, the first effective driving data and the second effective driving data are respectively subjected to probability density fitting to obtain the probability density fitting curve representing the probability density distribution of the first effective driving data and the probability density fitting curve representing the probability density distribution of the second effective driving data; S5, based on JS divergence, the JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data is obtained, and the JS divergence value is used to characterize the degree of difference between driving behavior in plateau environment and driving behavior in plain environment.

[0006] In some embodiments of the present invention, the constant altitude plateau environment is characterized by a constant oxygen concentration at a constant altitude, and the oxygen concentration is kept constant by adjusting a hypoxia generator; the plain driving data and the plateau driving data are obtained by sampling at distance intervals at the total mileage of the plain driving data and the total mileage of the plateau driving data, respectively. The plain driving data and the plateau driving data are sample datasets including several sample data. Preferably, the distance sampling interval is 2m.

[0007] In some embodiments of the present invention, the number of subjects in the simulated real-vehicle driving test is not less than 2. Each subject performs a plain environment driving test and a plateau environment driving test, and obtains the total plain driving mileage, plain driving data, plateau driving mileage, and plateau driving data of each subject under a preset driving time. The preprocessing of the total plain driving mileage and plateau driving mileage includes the elimination of abnormal total driving mileage and the determination of the initial mileage range. The elimination of abnormal total driving mileage includes the elimination of the subject data with the shortest total plain driving mileage and / or plateau driving mileage among all subjects. The initial mileage range is [0, the minimum total plain driving mileage or plateau driving mileage among the remaining subjects]. The plain driving mileage and plateau driving mileage within the initial mileage range are the preprocessed plain driving mileage and the preprocessed plateau driving mileage, respectively.

[0008] In some embodiments of the present invention, the plain driving data and plateau driving data both include four data indicators: steering wheel rotation rate, lateral acceleration, distance from the center of the road, and speed.

[0009] In some embodiments of the present invention, the driving performance decline time window is obtained as follows: the preprocessed high-altitude driving mileage is divided into several mileage stages according to a preset time interval, and the standard deviation of the high-altitude driving data corresponding to each mileage stage is obtained; the standard deviation of any mileage stage is then quantified by the next-order difference using the following formula: , Where x represents any mileage stage, x+1 represents the next mileage stage, y(x) is the standard deviation of any mileage stage, y(x+1) is the standard deviation of the next mileage stage, and Δy x The first-order difference result is the forward difference of the standard deviation for any mileage stage; the first-order difference result Δy x A scatter plot is drawn from the scattered data, and the scatter data is fitted to obtain a logarithmic function fitting curve. The slope corresponding to each point on the logarithmic function fitting curve is obtained, and the slope within a preset slope interval is selected from the slopes. The stage interval corresponding to the slope within the preset slope interval is obtained as the invalid stage interval. The time interval corresponding to the invalid stage interval is obtained based on the preset time interval, and the time interval is the time window of declining driving performance.

[0010] In some embodiments of the present invention, the mileage value corresponding to the maximum value of the time interval is recorded as the minimum effective mileage value, and the effective mileage interval is [the mileage value corresponding to the maximum value of the time interval, and the preprocessed high-altitude driving mileage endpoint value].

[0011] In some embodiments of the present invention, the specific process of step S4 is as follows: Based on the nonparametric kernel density estimation method, the probability density of the first effective driving data is fitted using the following probability density function: ,in, Here, n1 is the probability density function used to fit the probability density of the first valid driving data, and h1 is the window width. For Gaussian kernel function, z is the currently sought sample data in the first valid driving data. i Let be the i-th sample data in the first effective driving data; based on the nonparametric kernel density estimation method, the probability density of the second effective driving data is fitted using the following probability density function: ,in, Here, n2 is the probability density function used to fit the probability density of the second effective driving data, and h2 is the window width. For Gaussian kernel function, , m is the currently sought sample data in the second valid driving data, m i This is the i-th sample data in the second set of valid driving data.

[0012] In some embodiments of the present invention, the specific process of step S5 is as follows: Based on the following formula for JS divergence, the JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data is obtained: , , ,in, denoted as JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data.

[0013] On the other hand, the present invention also provides a driving behavior analysis system for high-altitude environments. The system includes: a simulated real-vehicle driving environment construction and data acquisition unit, used to construct simulated real-vehicle driving plain environments and constant-altitude high-altitude environments, and acquire total plain driving mileage, plain driving data, total high-altitude driving mileage, and high-altitude driving data under a preset driving time; a preprocessing unit, used to preprocess the total plain driving mileage and total high-altitude driving mileage to obtain preprocessed plain driving mileage and preprocessed high-altitude driving mileage; and an effective driving data acquisition unit, used to obtain an effective mileage interval based on a driving performance decline time window under high-altitude environments, and record the driving data collected from the preprocessed high-altitude driving mileage within the effective mileage interval as the first effective driving data. According to the data, driving data collected from preprocessed plain driving mileage within the effective mileage range is recorded as the second effective driving data; a probability density fitting unit is used to perform probability density fitting on the first and second effective driving data respectively to obtain a probability density fitting curve representing the probability density distribution of the first effective driving data and a probability density fitting curve representing the probability density distribution of the second effective driving data; a JS divergence value acquisition unit is used to obtain the JS divergence value between the probability density distribution of the first and second effective driving data based on JS divergence, and the JS divergence value is used to characterize the degree of difference between driving behavior in plateau environment and driving behavior in plain environment; a control module is used to issue commands to control the execution of each unit of the system.

[0014] Furthermore, the present invention also provides an application of the driving behavior analysis method for plateau environments described in any of the above claims. Based on the driving behavior analysis method for plateau environments described in any of the above claims, the degree of difference between driving behavior in plateau environments and driving behavior in plain environments is obtained, and an auxiliary driving strategy is matched to driving in plateau environments based on the degree of difference.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) In the simulated real vehicle driving test, the present invention constructs a plain environment and a constant altitude plateau environment respectively, and obtains driving data in the plain environment and constant altitude plateau environment. Starting from the perspective of driving data rather than the biopsychological perspective, the JS divergence value is obtained by combining probability density fitting with JS divergence. The difference between driving behavior in the plateau environment and driving behavior in the plain environment is intuitively analyzed. Compared with the objective factors that exist in the real vehicle test and reduce the accuracy of driving behavior analysis in the plateau environment, the present invention avoids the interference of objective factors and has high accuracy in analyzing driving behavior in the plateau environment. Based on the difference between driving behavior in the plateau environment and driving behavior in the plain environment, the present invention can reasonably match auxiliary driving strategies for driving in the plateau environment and improve driving safety in the plateau environment.

[0016] (2) Drivers need a certain amount of time to adapt to the driving environment when they start driving, especially when the plains environment changes to the plateau environment. When they start driving, drivers often go through three stages: not sensing the change in the environment, sensing the change in the environment but not adapting, and adapting to the plateau environment. The driving data of the first two stages often cannot accurately reflect the actual driving behavior in the plateau environment. If they are used to analyze driving behavior in the plateau, the analysis results will be inaccurate. Therefore, this invention is based on the characteristics of the dramatic change in data indicators in the stage of sensing the change in the environment but not adapting. Based on the time window of the decline in driving performance in the plateau environment, the effective mileage range of fully adapting to the plateau environment is obtained. The driving behavior is analyzed based on the first effective driving data and the second effective driving data within the effective mileage range, and the analysis results are more accurate. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be described below.

[0018] Figure 1 This is a flowchart of a driving behavior analysis method for high-altitude environments according to an embodiment of the present invention; Figure 2 This is a graph showing the total driving mileage in plains and plateau areas for different subjects under a preset driving time, according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the scatter plot and fitting curve of the first-order difference results from different data indicators during the process of obtaining the time window of declining driving performance in a high-altitude environment, according to an embodiment of the present invention. Figure 3 (a) represents the steering wheel rotation rate angle. Figure 3 (b) represents the lateral acceleration angle. Figure 3 (c) represents the angle of offset from the center of the road; Figure 4This is a radar chart drawn from first effective driving data and second effective driving data from different data indicator perspectives, according to one embodiment of the present invention. Figure 4 (a) represents the steering wheel rotation rate angle. Figure 4 (b) represents the lateral acceleration angle. Figure 4 (c) represents the angle of offset from the road center. Figure 4 (d) represents the velocity angle; Figure 5 This is a comparison chart of the probability density fitting curves of the first effective driving data in a plateau environment and the second effective driving data in a plain environment for the steering wheel rotation rate angle, according to an embodiment of the present invention. Figure 6 This is a comparison of probability density fitting curves between two groups of first effective driving data in a plateau environment and second effective driving data in a plain environment, after being mixed and grouped in different ways according to an embodiment of the present invention; wherein, Figure 6 (a) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment is randomly divided into two groups. Figure 6 (b) is a comparison of the probability density fitting curves between the two groups when the second effective driving data in the plain environment is randomly divided into two groups. Figure 6 (c) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment and the second effective driving data in the plain environment are mixed and randomly divided into two groups. Figure 7 This is a comparison chart of the probability density fitting curves of the first effective driving data in a plateau environment and the second effective driving data in a plain environment for the lateral acceleration angle, according to an embodiment of the present invention. Figure 8 This is a comparison of probability density fitting curves between two groups of first effective driving data in a plateau environment and second effective driving data in a plain environment, after being mixed and grouped in different ways according to an embodiment of the present invention; wherein, Figure 8 (a) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment is randomly divided into two groups. Figure 8 (b) is a comparison of the probability density fitting curves between the two groups when the second effective driving data in the plain environment is randomly divided into two groups. Figure 8 (c) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment and the second effective driving data in the plain environment are mixed and randomly divided into two groups. Figure 9 This is a comparison chart of the probability density fitting curves of the first effective driving data in a plateau environment and the second effective driving data in a plain environment for the offset road center distance angle according to an embodiment of the present invention. Figure 10 This is a comparison of probability density fitting curves between two groups of first effective driving data in a plateau environment and second effective driving data in a plain environment, after being mixed and grouped in different ways according to an embodiment of the present invention; wherein, Figure 10 (a) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment is randomly divided into two groups. Figure 10 (b) is a comparison of the probability density fitting curves between the two groups when the second effective driving data in the plain environment is randomly divided into two groups. Figure 10 (c) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment and the second effective driving data in the plain environment are mixed and randomly divided into two groups. Figure 11 This is a comparison chart of the probability density fitting curves of the first effective driving data in a plateau environment and the second effective driving data in a plain environment for speed angle, according to an embodiment of the present invention. Figure 12 This is a comparison of probability density fitting curves between two groups of speed and angle data from a high-altitude environment and a low-altitude environment, respectively, after being mixed and grouped in different ways according to an embodiment of the present invention; wherein, Figure 12 (a) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment is randomly divided into two groups. Figure 12 (b) is a comparison of the probability density fitting curves between the two groups when the second effective driving data in the plain environment is randomly divided into two groups. Figure 12 (c) is a comparison of the probability density fitting curves between the two groups when the first effective driving data in the plateau environment and the second effective driving data in the plain environment are mixed and randomly divided into two groups. Detailed Implementation

[0019] To make the objectives, technical solutions and advantages of the present invention clearer, the various aspects of the present invention will be described in detail below with reference to specific embodiments. However, these specific embodiments are only used to illustrate the present invention and do not constitute any limitation on the scope of protection and the substantive content of the present invention.

[0020] This embodiment provides a method for analyzing driving behavior in high-altitude environments, such as... Figure 1As shown, the process includes the following steps: S1. Based on simulated real-vehicle driving tests, a plain environment and a constant-altitude plateau environment are constructed respectively, and the total mileage, plain driving data, plateau driving mileage, and plateau driving data of the preset driving time are obtained; S2. The total mileage of the plain driving and the total mileage of the plateau driving are preprocessed to obtain the preprocessed plain driving mileage and the preprocessed plateau driving mileage; S3. Based on the time window of driving performance decline in the plateau environment, the effective mileage interval is obtained. The driving data collected in the preprocessed plateau driving mileage within the effective mileage interval is recorded as the first effective driving data, and the driving data collected in the preprocessed plain driving mileage within the effective mileage interval is recorded as the second effective driving data; S4. The probability density is fitted to the first effective driving data and the second effective driving data respectively to obtain the probability density fitting curve representing the probability density distribution of the first effective driving data and the probability density fitting curve representing the probability density distribution of the second effective driving data; S5. Based on JS divergence, the JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data is obtained. The JS divergence value is used to characterize the degree of difference between driving behavior in the plateau environment and driving behavior in the plain environment.

[0021] In this embodiment, the simulated vehicle driving test hardware system consists of a simulated vehicle cockpit, a circular display screen, an audio-visual system, and a power supply system, along with UC-winRoad 3D simulation software. This system enables panoramic modeling of the road driving environment and takeover scenarios, and real-time recording of vehicle operation and driving data. In this embodiment, the constant-altitude plateau environment is characterized by a constant oxygen concentration at a constant altitude. This oxygen concentration is maintained constant by adjusting a hypoxic generator. Specifically, the Everest Summit II hypoxic generator developed by Hypoxico can be used. By increasing the device's setting, the absolute oxygen content inside the mask worn by the subject is reduced, simulating a constant-altitude plateau environment in a plain area. Since the experiment is conducted in a plain area, the construction of the plain environment (i.e., the normal environment) is the actual natural environment without the hypoxic generator activated.

[0022] In this embodiment, the number of subjects in the simulated real-vehicle driving test is no less than 2. Each subject undergoes both a plain driving test and a high-altitude driving test, and the total mileage, data, and distance traveled during the plain driving test and high-altitude driving test are obtained for each subject under preset driving time. In this embodiment, taking 24 subjects as an example, all subjects are required to be male drivers with a Class C driver's license who have no experience driving or living in high-altitude environments, no respiratory diseases such as asthma, an average age of 23.92 years (age standard deviation 1.02 years), an average driving experience of 2.79 years (driving experience standard deviation 1.32 years), and to avoid the influence of other physical factors on driving behavior, subjects are required to be in good health before the test and avoid staying up late or drinking alcohol. Each subject's trial consisted of two independent simulated driving sessions, corresponding to driving in a plains environment and driving in a high-altitude environment, respectively. To protect subjects entering a high-altitude environment for the first time, the duration of each simulated driving session was preferably set to 20 minutes, i.e., the preset driving time was 20 minutes. To reduce the negative effects of prolonged driving, subjects were required to leave the simulator for a 10-minute rest when switching driving environments. To eliminate the sequence effect, the order of the driving environments was balanced among subjects, and the order of the plains and high-altitude driving types was shuffled. To avoid interference from events, no other driving tasks or secondary driving tasks were set during the driving process. In this embodiment, the test road was designed with a speed limit of 120 km / h. -1 The cross-section is a two-way six-lane road with a lane width of 3.75m and a green median. During the experiment, subjects were required to drive in the rightmost lane. The left side of the vehicle was marked by a white dashed line, indicating a left turn to another lane, while the right side was marked by a white solid line, indicating a right turn to the emergency lane. The available space on the left side of the vehicle was significantly greater than on the right. To eliminate interference from other vehicles in random traffic flow, no random traffic flow was generated during the test. In this embodiment, the constant altitude plateau environment can be any altitude, selected as needed. This embodiment uses an altitude of 3900m as an example, selecting the 11th setting of the hypoxic generator to match the actual 3900m altitude. During the preparation phase, subjects wore masks and adjusted their positions in the simulated driving cabin for approximately 5 minutes of simulated driving practice to adapt to the hypoxic generator, mask, and driving simulator. In the formal test, subjects were required to adhere to the speed limit and remain in their current lane as much as possible. Total mileage driven on flat ground, total mileage driven on plateau, and total mileage driven on plateau were collected for each subject during the formal test.

[0023] In this embodiment, the plain driving data and plateau driving data for each subject are obtained by sampling at distance intervals, respectively, based on the total plain driving mileage and the total plateau driving mileage. Each subject's plain driving data and plateau driving data are sample datasets containing several sample data points. Preferably, the distance sampling interval is 2 meters. In this embodiment, the distance interval sampling method can reduce the linear influence of driving at different locations on the road at the same time. Selecting a distance sampling interval of 2 meters can maximize the retention of more original data and restore the authenticity of the data. In this embodiment, since there are 24 subjects, each subject corresponds to one total plain driving mileage, one plain driving data point, one total plateau driving mileage, and one plateau driving data point, that is, a total of 24 total plain driving mileage, 24 plain driving data points, 24 total plateau driving mileage, and 24 plateau driving data points are included, and each plain driving data point and plateau driving data point is a sample dataset containing several sample data points. Figure 2 As shown, the total driving mileage at low altitudes and high altitudes for each of the 24 subjects (corresponding to numbers 1-24) is displayed. Since the total driving mileage at low altitudes or high altitudes varies among the subjects, this embodiment preprocesses the total driving mileage at low altitudes or high altitudes. The preprocessing includes removing abnormal driving mileage and determining the initial mileage range. The removal of abnormal driving mileage includes removing data from subjects with the shortest total driving mileage at low altitudes and / or high altitudes (here, "shortest" indicates an abnormal mileage compared to other data). Figure 2 As shown, compared to other subjects, subject number 24 had a shorter total plain driving mileage and total plateau driving mileage, therefore subject number 24's data was excluded. Since the total plain driving mileage or total plateau driving mileage varies among subjects, excessively long mileages need to be processed to obtain an initial mileage interval. This initial mileage interval is defined as [0, the minimum total plain driving mileage or total plateau driving mileage among the remaining subjects]. The plain driving mileage and plateau driving mileage within the initial mileage interval are respectively the preprocessed plain driving mileage and the preprocessed plateau driving mileage. In this embodiment, as... Figure 2 As shown, among the remaining subjects, subject number 8 had the smallest total driving mileage on plains and on plateaus, at 29760m. Therefore, the initial mileage interval is [0, 29760m].

[0024] In this embodiment, both plain driving data and plateau driving data include four data indicators: steering wheel rotation rate, lateral acceleration, distance from the road center, and speed. Data collection requires acquiring driving data for all four indicators. In this embodiment, each subject corresponds to one set of driving data for plain environment (steering wheel rotation rate, lateral acceleration, distance from the road center, and speed), and one set of driving data for plateau environment (steering wheel rotation rate, lateral acceleration, distance from the road center, and speed). Each data set is a sample dataset containing several sample data points. In this embodiment, the distance from the road center is the distance from the center of the vehicle's front to the center of the entire road.

[0025] In this embodiment, the driving performance decline time window is obtained as follows: the preprocessed high-altitude driving mileage is divided into several mileage stages according to a preset time interval, and the standard deviation of the high-altitude driving data corresponding to each mileage stage is obtained; the standard deviation of any mileage stage is then quantified by the next-order difference using the following formula: , Where x represents any mileage stage, x+1 represents the next mileage stage, y(x) is the standard deviation of any mileage stage, y(x+1) is the standard deviation of the next mileage stage, and Δy x The first-order difference result is the forward difference of the standard deviation for any mileage stage; the first-order difference result Δy xA scatter plot is generated from the scattered data, and a logarithmic function fitting curve is obtained. The slope corresponding to each point on the logarithmic function fitting curve is obtained, and the slope within a preset slope interval is selected. The stage interval corresponding to the slope within the preset slope interval is obtained as the invalid stage interval. The time interval corresponding to the invalid stage interval is obtained based on a preset time interval, and the time interval is the driving performance decline time window. In this embodiment, the mileage value corresponding to the maximum value of the time interval is recorded as the minimum effective mileage value, and the effective mileage interval is [the mileage value corresponding to the maximum value of the time interval, and the preprocessed high-altitude driving mileage endpoint value]. In this embodiment, the driving performance decline time window is obtained based on three data indicators: steering wheel rotation rate, lateral acceleration, and distance from the center of the road. Taking steering wheel rotation rate as an example, with a preset time interval of 8.5 seconds, the entire 29760m journey was divided into 120 mileage stages based on the average vehicle speed (105km / h). Each subject (the remaining 23) had a steering wheel rotation rate sample dataset corresponding to several steering wheel rotation rate samples at any given mileage stage, based on the number of samples taken in that stage. The steering wheel rotation rate sample datasets of all subjects in the same mileage stage were merged into one single dataset. The standard deviation of the steering wheel rotation rate samples in the merged dataset across the 120 mileage stages was calculated, resulting in 120 standard deviations of steering wheel rotation rate data. These standard deviations were then calculated using the formula... For any mileage stage, the standard deviation is quantified by the first difference, and the first difference result Δy is... x The data is plotted as a scatter plot, and the scatter plot is fitted to obtain a logarithmic function fitting curve, such as... Figure 3 As shown in (a), the scatter plot and fitted curve of the first-order difference result for the steering wheel rotation rate angle are presented. Similarly, performing the same operation as the steering wheel rotation rate on the lateral acceleration and the distance from the road center, as shown in (a), yields the following results: Figure 3 (b) and Figure 3 The first-order difference results of the lateral acceleration angle (c) and the fitted curve are shown in the figure. Specifically, the fitted curve for the steering wheel rotation rate angle is: y steeringV =50.60ln(x)+314.23, the fitted curve for the lateral acceleration angle is: The fitted curve for the angle of offset from the road center is: After obtaining the fitted curves, the slope corresponding to each point on each fitted curve is obtained. The slopes that fall within the preset slope interval (defined as the interval composed of slopes whose slope values ​​change significantly among all the slopes of each fitted curve) are selected from the slopes. In this embodiment, the stage intervals corresponding to the slopes within the preset slope intervals of the three fitted curves are all [30, 40], that is, the invalid stage interval is [30, 40], the corresponding time interval is [255s, 340s], that is, the driving performance decline time window is [4.25min, 5.67min], and the effective mileage interval is [the mileage value corresponding to driving at an average speed for 5.67min, 29760m]. Based on this, the first set of valid driving data includes a sample dataset composed of several sample data points collected in a high-altitude environment within 23 valid mileage intervals [mileage value corresponding to 5.67 minutes of driving at an average speed, 29760m]. The second set of valid driving data includes a sample dataset composed of several sample data points collected in a plain environment within 23 valid mileage intervals [mileage value corresponding to 5.67 minutes of driving at an average speed, 29760m]. Both the first and second sets of valid driving data cover driving data based on four indicators: steering wheel rotation angle, lateral acceleration angle, distance from the center of the road offset angle, and speed angle. Specifically, the first set of valid driving data is divided into first set of valid driving data for steering wheel rotation angle, lateral acceleration angle, distance from the center of the road offset angle, and speed angle. This includes a steering wheel rotation angle sample dataset composed of several sample data points collected in a high-altitude environment within 23 valid mileage intervals [mileage value corresponding to 5.67 minutes of driving at an average speed, 29760m], and a lateral acceleration angle sample dataset composed of several sample data points collected in a plain environment within 23 valid mileage intervals [mileage value corresponding to 5.67 minutes of driving at an average speed, 29760m]. The first set of effective driving data consists of several lateral acceleration angle sample datasets collected within a mileage range of 29760m in the plateau environment; two sets of effective driving data consists of several speed angle sample datasets collected within a mileage range of 23 offset road center distance angles [based on an average driving speed of 5.67min, 29760m] in the plateau environment; and two sets of effective driving data consists of several speed angle sample datasets collected within a mileage range of 23 offset road center distance angles [based on an average driving speed of 5.67min, 29760m]. The second set of effective driving data includes steering wheel rotation angle, lateral acceleration angle, offset road center distance angle, and speed angle. The second set of effective driving data includes steering wheel rotation angle sample datasets collected within a mileage range of 23 offset road center distance angles [based on an average driving speed of 5.67min, 29760m] in the plain environment; and two sets of effective driving data consists of steering wheel rotation angle sample datasets collected within a mileage range of 23 offset road center distance angles [based on an average driving speed of 5.67min, 29760m].The dataset comprises a lateral acceleration angle sample dataset (29760m) collected within a mileage interval corresponding to 67 minutes of driving time; an offset road center distance angle sample dataset (29760m) collected within a mileage interval corresponding to 23 offset road center distance angles; and a velocity angle sample dataset (29760m) collected within a mileage interval corresponding to 5.67 minutes of driving time at an average speed.

[0026] like Figure 4 As shown, a radar chart is plotted from different data indicators using first and second effective driving data. Taking steering wheel rotation rate as an example, the standard deviation of several steering wheel rotation rate sample data collected in a plateau environment within the effective mileage interval of each subject's steering wheel rotation rate angle [the mileage value corresponding to 5.67 minutes of driving at an average speed, 29760m] is calculated, resulting in 23 standard deviation data points for steering wheel rotation rate. Similarly, the standard deviation of several steering wheel rotation rate sample data collected in a plain environment within the effective mileage interval of each subject's steering wheel rotation rate angle [the mileage value corresponding to 5.67 minutes of driving at an average speed, 29760m] is calculated, resulting in 23 standard deviation data points. The resulting chart is shown below. Figure 4 The curve comparison diagram shown in (a) is as follows. Similarly, by performing the same operations as the steering wheel rotation rate on lateral acceleration, distance from the road center, and velocity, the following results are obtained: Figure 4 (b) Figure 4 (c) and Figure 4 A curve comparison chart of (d). From Figure 4 It can be seen that, compared with the plains environment, in the plateau environment, most of the test subjects' ability to control the steering wheel decreased, resulting in fluctuations in lateral acceleration, causing the vehicle to frequently deviate from its current position, and the ability to control speed was weaker, with larger fluctuations in vehicle speed.

[0027] In this embodiment, the specific process of step S4 is as follows: Based on the nonparametric kernel density estimation method, the probability density of the first effective driving data is fitted using the following probability density function: ,in, Here, n1 is the probability density function used to fit the probability density of the first valid driving data, and h1 is the window width, which is generally chosen based on minimizing the mean square error. For Gaussian kernel function, z is the currently sought sample data in the first valid driving data. i Let be the i-th sample data in the first effective driving data; based on the nonparametric kernel density estimation method, the probability density of the second effective driving data is fitted using the following probability density function: ,in, Here, n² is the probability density function used to fit the probability density of the second effective driving data, and h² is the window width, typically chosen based on minimizing the mean square error. For Gaussian kernel function, , m is the currently sought sample data in the second valid driving data, m i This is the i-th sample data in the second effective driving data. In this embodiment, taking steering wheel rotation rate as an example, the steering wheel rotation rate sample dataset, composed of several steering wheel rotation rate sample data collected in the plateau environment within the effective mileage interval of the 23 steering wheel rotation rate angles of the first effective driving data [the mileage value corresponding to driving at an average speed of 5.67 minutes, 29760m], is merged into one sample dataset. Similarly, the steering wheel rotation rate sample dataset, composed of several steering wheel rotation rate sample data collected in the plain environment within the effective mileage interval of the 23 steering wheel rotation rate angles of the second effective driving data [the mileage value corresponding to driving at an average speed of 5.67 minutes, 29760m], is merged into one sample dataset. Then, the probability density fitting in step S4 above is performed, where n1 is the total number of sample data in the merged sample dataset of the first effective driving data, z is the currently sought sample data in the merged sample dataset of the first effective driving data, and z... i Let n1 be the i-th sample data in the merged sample dataset of the first valid driving data, n2 be the total number of sample data in the merged sample dataset of the second valid driving data, and m be the currently sought sample data in the merged sample dataset of the second valid driving data. i This refers to the i-th sample data in the merged sample dataset of the second set of valid driving data. Similarly, the same operations as those applied to lateral acceleration, distance from the road center, and speed are performed on these parameters. Figure 5 , 7 Figures 9 and 11 show comparison charts of the probability density fitting curves for the first effective driving data in a plateau environment and the second effective driving data in a plain environment, respectively, for steering wheel rotation rate angle, lateral acceleration angle, distance from road center angle, and speed angle. Figure 5 It can be seen that the peak value of the right-hand (positive) steering wheel rotation rate decreases slightly in high-altitude environments, indicating that it is easier to turn the steering wheel to the right in high-altitude environments, and the driver's lateral control over the vehicle decreases. Figure 7 It can be seen that the rightward (positive) acceleration value is larger in high-altitude environments, indicating that vehicles are more likely to generate rightward acceleration in high-altitude environments, and that the acceleration value is also larger. Figure 9 It can be seen that the peak value at 0m is approximately 0.8 in a plain environment, while it is lower at the same location in a plateau environment, approximately 0.6. The fluctuation range in the plain environment is from -7m to 2.5m, while the fluctuation range in the plateau environment is wider, between -8m and 4m. This indicates that in a plain environment, vehicles generally maintain lateral stability and rarely cross the solid white line to the right to reach the emergency lane. However, in a plateau environment, vehicles cannot maintain a stable lateral position on the current road, and the number of times they deviate to the right increases, while the distances of deviation to the right are also greater. Figure 11 It can be seen that the peak value in the plain environment appears at about 108 km / h, while the peak value in the plateau environment shifts significantly to the right, appearing at about 114 km / h. The peak value of the probability density in the plateau environment is about 0.038, which is higher than 0.034 in the plain environment. In the plateau environment, the area of ​​the probability density fitting curve exceeding the speed limit of 120 km / h is larger, indicating that the vehicle speed generally increases in the plateau environment, and the speeding phenomenon is more serious.

[0028] In this embodiment, the data is considered authentic only when the driving behavior of the subjects in the same environment tends to be consistent. Otherwise, the analysis based on the driving data of 23 subjects will be different from reality and cannot be used to represent real driving behavior, let alone as a basis for matching assisted driving strategies. Therefore, in this embodiment, the first effective driving data and the second effective driving data are mixed in different ways and divided into two groups. The authenticity of the driving data is verified by the probability density fitting curve between the two groups. Specifically, taking steering wheel rotation rate as an example, the data is grouped in the following way (as shown in Table 1): ① The steering wheel rotation rate angle sample dataset, consisting of several steering wheel rotation rate angle sample data collected in the plateau environment within the effective mileage interval of the 23 steering wheel rotation rate angles of the first effective driving data [mileage value corresponding to driving at an average speed of 5.67 min, 29760 m], is merged into one sample dataset. The sample data in the merged sample dataset is randomly divided into two groups to obtain two data sample sets, denoted as the plateau environment grouping (D-1, D-2); ② The steering wheel rotation rate angle sample dataset, consisting of several steering wheel rotation rate angle sample data collected in the plain environment within the effective mileage interval of the 23 steering wheel rotation rate angles of the second effective driving data [mileage value corresponding to driving at an average speed of 5.67 min, 29760 m], is merged into one sample dataset. The sample data in the merged sample dataset is randomly divided into two groups to obtain two data sets. The sample sets are denoted as the plain environment group (P-1, P-2); ③ The steering wheel rotation angle sample dataset, which consists of several steering wheel rotation angle sample data collected in the plateau environment within the effective mileage interval of the 23 steering wheel rotation angles of the first effective driving data [mileage value corresponding to driving at an average speed of 5.67 min, 29760 m], is merged into one sample dataset. The steering wheel rotation angle sample dataset, which consists of several steering wheel rotation angle sample data collected in the plain environment within the effective mileage interval of the 23 steering wheel rotation angles of the second effective driving data [mileage value corresponding to driving at an average speed of 5.67 min, 29760 m], is merged into one sample dataset. The two merged sample datasets are further merged, and the sample data in them are randomly divided into two groups to obtain two data sample sets, denoted as the plain environment + plateau environment mixed group (preferably, the mixed group of D-1 and P-1, and the mixed group of D-2 and P-2). Then, probability density fitting is performed on the grouped D-1 and D-2 separately, and on the grouped P-1 and P-2 separately. Probability density fitting is also performed on the mixed groups of D-1 and P-1, and D-2 and P-2 separately. Similarly, the same operations as those performed on steering wheel rotation rate are applied to lateral acceleration, distance from the road center, and velocity. Figure 6 , 8Figures 10 and 12 show a comparison of probability density fitting curves obtained using the three grouping methods ①②③ mentioned above, for steering wheel rotation rate angle, lateral acceleration angle, distance from road center angle, and velocity angle. Figure 6 It can be seen that, regardless of the data indicator used—whether it's D-1 and D-2 within the plateau environment group, P-1 and P-2 within the plain environment group, or the mixed group of D-1 and P-1 and the mixed group of D-2 and P-2 in both plain and plateau environments—the probability density fitting curves basically overlap. Furthermore, the steering wheel rotation angle and lateral acceleration angle exhibit double peaks symmetrical about the zero axis; the positions of the peaks are essentially the same across different scenarios. Therefore, the driving behaviors of the 23 subjects in this embodiment under the same environment (comparison within the plateau group, comparison within the plain group, and comparison of the mixed plain and plateau environments) tend to be consistent, demonstrating authenticity. This data can be used to characterize real driving behavior and can serve as a basis for matching assisted driving strategies, thereby improving driving safety in plateau environments.

[0029] Table 1. Grouping of First Effective Driving Data in Plateau Environment and Second Effective Driving Data in Plain Environment In this embodiment, the specific process of step S5 is as follows: Based on the following formula for JS divergence, the JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data is obtained: , , ,in, The JS divergence value is the difference between the probability density distributions of the first and second effective driving data. In this embodiment, the JS divergence value is calculated from four data indicators: steering wheel rotation angle, lateral acceleration angle, distance from road center angle, and speed angle. The JS divergence value ranges from 0 to 1. The larger the JS divergence value (the closer it is to 1), the greater the difference between the two probability distributions and the lower the similarity. Table 2 below shows the JS divergence values ​​obtained using the driving data of 23 subjects selected in this embodiment as an example. Table 2 also includes the JS divergence values ​​obtained from the corresponding probability density distributions when grouped using the three grouping methods ①②③ mentioned above.

[0030] Table 2. JS divergence values ​​between probability density distributions under different grouping methods As shown in Table 2, the JS divergence values ​​obtained from the probability density distributions of the four data indicators—steering wheel rotation rate, lateral acceleration, distance from the center of the road, and speed—are all 0, based on the grouping methods of plain environment intra-group (P-1, P-2), plateau environment intra-group (D-1, D-2), and mixed plain and plateau environment group (D-1+P-1, D-2+P-2). This indicates that there is no difference between the two probability distributions, and they are completely similar. This further demonstrates that the driving behaviors of the 23 subjects in this embodiment tend to be consistent in the same environment (comparison within the plateau group, comparison within the plain group, and comparison of the mixed plain and plateau environment), which is authentic and can be used to characterize real driving behavior. It can also be used as a basis for matching assisted driving strategies to improve driving safety in plateau environments. The JS divergence values ​​obtained from the probability density distributions of the plateau environment group and the plain environment group are significantly higher than those of other grouping methods. This indicates that the probability density distributions of road center deviation distance, lateral acceleration, steering wheel rotation rate, and speed in the plateau environment are most significantly different from those in the plain environment, with JS divergence values ​​of 0.23, 0.11, 0.01, and 0.02, respectively. Except for the steering wheel rotation rate, the JS divergence values ​​of the lateral indicators lateral acceleration and road center deviation distance are much greater than those of the longitudinal indicator speed. This indicates that the lateral stability of the vehicle decreases more significantly in the plateau environment. In the plateau environment, the degree of difference between driving behavior in the plateau environment and driving behavior in the plain environment (as opposed to similarity, the greater the similarity, the smaller the difference) quantified by the JS divergence value in this embodiment can be used as a basis to start from the significant decrease in vehicle lateral stability. Based on the magnitude of the difference, different assisted driving strategies can be matched to improve driving safety in the plateau environment.

[0031] This embodiment also provides a driving behavior analysis system for high-altitude environments, including: a simulated real-vehicle driving environment construction and data acquisition unit, used to construct a simulated real-vehicle driving plain environment and a constant altitude high-altitude environment, and acquire the total plain driving mileage, plain driving data, total high-altitude driving mileage, and high-altitude driving data under a preset driving time; a preprocessing unit, used to preprocess the total plain driving mileage and total high-altitude driving mileage to obtain preprocessed plain driving mileage and preprocessed high-altitude driving mileage; and an effective driving data acquisition unit, used to obtain an effective mileage interval based on the driving performance decline time window under high-altitude environment, and record the driving data collected in the preprocessed high-altitude driving mileage within the effective mileage interval as the first effective driving data, and then... Driving data collected within the preprocessed plain driving mileage range within the effective mileage interval is recorded as the second effective driving data; a probability density fitting unit is used to perform probability density fitting on the first and second effective driving data respectively to obtain a probability density fitting curve representing the probability density distribution of the first effective driving data and a probability density fitting curve representing the probability density distribution of the second effective driving data; a JS divergence value acquisition unit is used to obtain the JS divergence value between the probability density distribution of the first and second effective driving data based on JS divergence, and to characterize the degree of difference between driving behavior in plateau environment and driving behavior in plain environment through the JS divergence value; a control module is used to issue commands to control the execution of each unit of the system.

[0032] This embodiment also provides the application of the driving behavior analysis method for plateau environments. Based on the driving behavior analysis method for plateau environments in this embodiment, the degree of difference between driving behavior in plateau environments and driving behavior in plain environments is obtained, and assisted driving strategies are matched to driving in plateau environments based on the degree of difference.

[0033] The present invention has been described above with reference to specific embodiments. These specific embodiments are merely exemplary and should not be construed as limiting the scope of protection of the present invention. Those skilled in the art can make various modifications, changes, or substitutions without departing from the essence of the present invention. Therefore, various equivalent variations made according to the present invention still fall within the scope of the present invention.

Claims

1. A driving behavior analysis method for a highland environment, characterized by, Includes the following steps: S1. Based on simulated real vehicle driving tests, a plain environment and a constant altitude plateau environment are constructed respectively, and the total mileage of plain driving, plain driving data, total mileage of plateau driving, and plateau driving data are obtained under the preset driving time. S2. Preprocess the total driving mileage in plains and the total driving mileage in plateau areas to obtain the preprocessed driving mileage in plains and the preprocessed driving mileage in plateau areas. S3. Based on the time window of declining driving performance in high-altitude environments, the effective mileage range is obtained. The driving data collected from the preprocessed high-altitude driving mileage within the effective mileage range is recorded as the first effective driving data, and the driving data collected from the preprocessed plain driving mileage within the effective mileage range is recorded as the second effective driving data. S4. Perform probability density fitting on the first effective driving data and the second effective driving data respectively to obtain the probability density fitting curve representing the probability density distribution of the first effective driving data and the probability density fitting curve representing the probability density distribution of the second effective driving data. S5. Based on JS divergence, obtain the JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data. The JS divergence value is used to characterize the degree of difference between driving behavior in a plateau environment and driving behavior in a plain environment. The method for obtaining the driving performance decline time window is as follows: The preprocessed high-altitude driving mileage is divided into several mileage stages according to a preset time interval, and the standard deviation of the high-altitude driving data corresponding to each mileage stage is obtained. The standard deviation of any mileage stage is quantified to the first-order difference using the following formula: , , where x is any mileage stage, x+1 is the next mileage stage, y(x) is the standard deviation of any mileage stage, y(x+1) is the standard deviation of the next mileage stage, and Δyx is the first difference result of the forward difference of the standard deviation of any mileage stage. The first-order difference result Δyx is plotted as a scatter plot in the form of scattered data. The scatter data is then fitted to obtain the logarithmic function fitting curve. Obtain the slope corresponding to each point on the logarithmic function fitting curve, select the slope within a preset slope interval from the slopes, and obtain the stage interval corresponding to the slope within the preset slope interval as the invalid stage interval; The time interval corresponding to the invalid phase interval is obtained based on the preset time interval, and the time interval is the time window for the decline in driving performance. The mileage value corresponding to the maximum value of the time interval is recorded as the minimum effective mileage value, and the effective mileage interval is [the mileage value corresponding to the maximum value of the time interval, and the preprocessed plateau driving mileage endpoint value].

2. The driving behavior analysis method for high-altitude environments as described in claim 1, characterized in that, The constant-altitude plateau environment is characterized by a constant oxygen concentration at a constant altitude, and the oxygen concentration is kept constant by adjusting the hypoxia generator. Plain driving data and plateau driving data were obtained by sampling at distance intervals based on the total mileage of plain driving and the total mileage of plateau driving, respectively. The plain driving data and plateau driving data are sample datasets including several sample data, with a distance sampling interval of 2m.

3. The driving behavior analysis method for high-altitude environments as described in claim 1 or 2, characterized in that, The number of subjects in the simulated real vehicle driving test shall not be less than 2. Each subject shall conduct a driving test in a plain environment and a driving test in a high-altitude environment. The total mileage of plain driving, the total mileage of plain driving, the total mileage of high-altitude driving, and the high-altitude driving data of each subject shall be obtained under the preset driving time. The preprocessing of total driving mileage in plains and high-altitude areas includes the elimination of abnormal total driving mileage and the determination of the initial mileage range. The elimination of abnormal total driving mileage includes the removal of data from subjects with the shortest total driving mileage in plains and / or high-altitude areas among all subjects. The initial mileage range is [0, the minimum total driving mileage in plains or high-altitude areas among the remaining subjects]. The plain driving mileage and high-altitude driving mileage within the initial mileage range are the preprocessed plain driving mileage and the preprocessed high-altitude driving mileage, respectively.

4. The driving behavior analysis method for high-altitude environments as described in claim 1, characterized in that, Both the plain driving data and the plateau driving data include four data indicators: steering wheel rotation rate, lateral acceleration, distance from the center of the road, and speed.

5. The driving behavior analysis method for high-altitude environments as described in claim 1, characterized in that, The specific process of step S4 is as follows: Based on the nonparametric kernel density estimation method, the probability density of the first valid driving data is fitted using the following probability density function: ， in, Here, n1 is the probability density function used to fit the probability density of the first valid driving data, and h1 is the window width. For Gaussian kernel function, z is the currently sought sample data in the first valid driving data. i This is the i-th sample data in the first valid driving data; Based on the nonparametric kernel density estimation method, the probability density of the second effective driving data is fitted using the following probability density function: ， in, Here, n2 is the probability density function used to fit the probability density of the second effective driving data, and h2 is the window width. For Gaussian kernel function, , m is the currently sought sample data in the second valid driving data, m i This is the i-th sample data in the second set of valid driving data.

6. The driving behavior analysis method for high-altitude environments as described in claim 5, characterized in that, The specific process of step S5 is as follows: The JS divergence value between the probability density distributions of the first and second effective driving data is obtained using the following formula: ， ， ， in, denoted as JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data.

7. A driving behavior analysis system for high-altitude environments, characterized in that, The system includes: The simulated real vehicle driving environment construction and data acquisition unit is used to construct a simulated real vehicle driving plain environment and a constant altitude plateau environment, and to acquire the total plain driving mileage, plain driving data, total plateau driving mileage, and plateau driving data under a preset driving time. The preprocessing unit is used to preprocess the total driving mileage in plains and the total driving mileage in plateau areas to obtain the preprocessed driving mileage in plains and the preprocessed driving mileage in plateau areas. The effective driving data acquisition unit is used to obtain the effective mileage range based on the time window of driving performance decline in plateau environment. The driving data collected in the preprocessed plateau driving mileage within the effective mileage range is recorded as the first effective driving data, and the driving data collected in the preprocessed plain driving mileage within the effective mileage range is recorded as the second effective driving data. The probability density fitting unit is used to perform probability density fitting on the first effective driving data and the second effective driving data respectively, to obtain the probability density fitting curve representing the probability density distribution of the first effective driving data and the probability density fitting curve representing the probability density distribution of the second effective driving data. The JS divergence value acquisition unit is used to acquire the JS divergence value between the probability density distribution of the first effective driving data and the probability density distribution of the second effective driving data based on JS divergence, and to characterize the degree of difference between driving behavior in plateau environment and driving behavior in plain environment through the JS divergence value. The control module is used to issue commands to control the execution of various units within the system. The method for obtaining the driving performance decline time window is as follows: The preprocessed high-altitude driving mileage is divided into several mileage stages according to a preset time interval, and the standard deviation of the high-altitude driving data corresponding to each mileage stage is obtained. The standard deviation of any mileage stage is quantified to the first-order difference using the following formula: , Where x represents any mileage stage, x+1 represents the next mileage stage, y(x) is the standard deviation of any mileage stage, y(x+1) is the standard deviation of the next mileage stage, and Δy x The first-order difference result of the forward difference of the standard deviation for any mileage stage; The first-order difference result Δy x The scatter point data is plotted in a scatter plot, and the scatter point data is fitted to obtain a logarithmic function fitting curve; Obtain the slope corresponding to each point on the logarithmic function fitting curve, select the slope within a preset slope interval from the slopes, and obtain the stage interval corresponding to the slope within the preset slope interval as the invalid stage interval; The time interval corresponding to the invalid phase interval is obtained based on the preset time interval, and the time interval is the time window for the decline in driving performance. The mileage value corresponding to the maximum value of the time interval is recorded as the minimum effective mileage value, and the effective mileage interval is [the mileage value corresponding to the maximum value of the time interval, and the preprocessed plateau driving mileage endpoint value].

Images