A time-varying effect quantification method and device for a driving assistance warning system

Abstract

The application discloses a time-varying effect quantification method and equipment for a driving assistance warning system, and the method comprises the following steps: acquiring event-level running data and scene attribute data of a vehicle before and after triggering of the driving assistance warning system, and establishing an analysis time window with a warning triggering time as zero point; obtaining a pre-trigger trajectory and a post-trigger trajectory according to data in the analysis time window, and obtaining a net post-response effect curve after a control pre-response history and a pre-post transition coefficient surface according to the pre-trigger trajectory and the post-trigger trajectory; constructing a single-event no-warning counterfactual post-response curve, and obtaining a net response curve from a difference between the single-event no-warning counterfactual post-response curve and an actual post-trigger trajectory; clustering driver response types based on the net response curve, and obtaining different driver response types; setting differentiated warning thresholds according to statistical distribution of warning triggering indexes under the different driver response types, and dividing the warning into three categories. The application can quantize the effect of the driving assistance warning system on a continuous time, and supports adaptive threshold design for driver differences.

Description

Technical Field

[0001] This invention belongs to the fields of intelligent transportation, driving assistance, active safety warning and driving behavior analysis, and particularly relates to a method and device for quantifying the time-varying effect of a driving assistance warning system. Background Technology

[0002] With the development of advanced driver assistance systems (ADAS), vehicles can receive various safety alerts from onboard systems during operation, such as collision risk warnings, speeding warnings, cornering warnings, and following-risk warnings. The goal of driver assistance warning systems is to influence the driver's speed control and decision-making behavior through external prompts, thereby reducing risks and improving safety.

[0003] In existing technologies, the evaluation of the effectiveness of driver assistance warning systems mostly uses static scalar indicators such as average deceleration and average reaction time, which are difficult to characterize the dynamic evolution of the warning effect over continuous time. At the same time, existing methods fail to separate the behavioral preparation that exists before the trigger from the trajectory changes after the trigger, which can easily lead to an overestimation of the true effectiveness of the warning system. In addition, existing methods ignore the significant individual differences among different drivers in terms of response intensity, time delay, and degree of dependence on behavioral inertia.

[0004] In summary, the existing technologies have the following problems: First, existing methods mostly use scalar indicators such as average deceleration, average reaction time, or fixed time window difference, which are difficult to characterize the dynamic process of the driver assistance warning system over continuous time; Second, existing methods usually do not distinguish between the behavioral preparation formed before the warning is triggered and the behavioral correction added after the warning is triggered, which easily overestimates the true effect of the warning system; Third, existing methods have difficulty identifying the differences in response intensity, response lag, and persistence among different drivers, and therefore cannot support hierarchical and categorized warning threshold settings.

[0005] Therefore, there is a need for a method that can continuously quantify the effects of driver assistance warning systems, control the historical impact of pre-response, identify driver response types, and further perform threshold adaptive design, which leads to this case. Summary of the Invention

[0006] The purpose of this invention is to provide a method and device for quantifying the time-varying effect of a driving assistance warning system, which can quantify the effect of the driving assistance warning system over continuous time and support adaptive threshold design based on driver differences.

[0007] To achieve the above objectives, the solution of the present invention is:

[0008] A method for quantifying the time-varying effect of a driver assistance warning system includes the following steps:

[0009] Step 1: Obtain event-level operational data and scene attribute data of the vehicle before and after the driver assistance warning system is triggered, and establish an analysis time window with the warning trigger time as the zero point;

[0010] Step 2: Based on the data within the analysis time window, construct the normalized velocity change trajectory and divide it into the pre-trigger trajectory and the post-trigger trajectory;

[0011] Step 3: Establish a function response-scalar regression model based on the normalized velocity change trajectory of the full time window to obtain the direct time-varying effect curve of the driving assistance warning system, and extract the average effect before warning triggering, the average effect after warning triggering, and the moment when the warning effect first crosses zero.

[0012] Step 4: Based on the pre-trigger trajectory and the post-trigger trajectory, establish a function response-function regression model to obtain the net post-response effect curve and the front-to-back transition coefficient surface after controlling the pre-response history; further construct a single-event no-warning counterfactual post-response curve, and obtain the net response curve from the difference between the actual post-trigger trajectory and the no-warning counterfactual post-response curve;

[0013] Step 5: Extract the pre-response intensity, net post-response intensity, response lag, peak response, effective duration and pre- and post-coupling degree indicators based on the net response curve, and cluster the driver response types based on the indicators to obtain different driver response types.

[0014] Step 6: Set differentiated alarm thresholds based on the statistical distribution of alarm triggering indicators under different driver response types, and divide alarms into three categories. The alarm thresholds for the three categories are respectively set as the alarm warning mean minus the standard deviation, the alarm warning mean, and the alarm warning mean plus the standard deviation.

[0015] In step 1 above, the event-level operational data includes event identifier, timestamp, vehicle speed, vehicle location, and warning issuance marker, while the scene attribute data includes at least one of the following: time period identifier, road type, traffic status, speed limit information, and road environment attributes.

[0016] The specific content of step 2 above is as follows:

[0017] Step 21: Based on the data within the analysis time window, perform B-spline smoothing reconstruction on the triggered aligned discrete velocity trajectory to obtain the first... Continuous velocity trajectory function of an event ,in, Indicates the event number. A continuous time variable representing the relative warning trigger time. Indicates the first The event at time Continuous velocity;

[0018] Step 22, construct the normalized velocity change trajectory.

[0019]

[0020] in, Indicates the first The baseline velocity of an event at the left end of the analysis time window. Indicates the first The velocity change trajectory of an event relative to a reference velocity; This indicates the half-length of the analysis time window;

[0021] Step 23: Divide the normalized velocity change trajectory into a pre-trigger trajectory and a post-trigger trajectory.

[0022]

[0023]

[0024] in, This indicates the trajectory of speed changes before the warning is triggered. This indicates the trajectory of speed changes after the warning is triggered.

[0025] In step 21 above, let the first... The discrete velocity observations of each event within the analysis time window are as follows: ,in, Indicates the first The observation point number in each event, Indicates the first The total number of observation points for each event. Indicates the first Time coordinates of each observation point Indicates the first Velocity observation values ​​at each observation point;

[0026] Assume that discrete observations satisfy,

[0027]

[0028] in, Represents the observation error term; take indivual The basis functions of the second-order B-spline form the basis function vector.

[0029]

[0030] in, Indicates the first indivual B-spline basis functions , Indicates transpose;

[0031] The continuous velocity trajectory function is expressed as follows:

[0033] in, Indicates the first The spline coefficient vector of events; let

[0034]

[0035] in, Represents the velocity observation vector. Represents the spline design matrix;

[0036] Solve

[0037]

[0038] The spline coefficient estimates are obtained, where, express 3D real space, Represents the L2 norm, Indicates the first Smoothing parameters for each event, Represents the roughness penalty matrix;

[0039] The corresponding reconstruction speed trajectory is

[0040]

[0041] Smoothing parameters Determined through generalized cross-validation

[0042]

[0043] in, Represents candidate smoothing parameters The generalized cross-validation value under the following conditions The parameter is Spline coefficient estimation at time, Represents the trace of a matrix.

[0044]

[0045] This represents a smoothing matrix.

[0046] In step 3 above, the function response-scalar regression model is expressed as follows:

[0047]

[0048] in, This represents the reference response function under reference conditions. This represents the direct time-varying effect function of the warning issuance on the normalized velocity change trajectory. The time-varying influence function representing the baseline velocity, The time-varying coefficient vector representing the scene covariate vector. Represents the residual function; This indicates that a warning has been issued. This indicates that a warning has been issued. This indicates that no warning was issued or a silent comparison was performed; Indicates the first The baseline velocity of each event, Indicates the first an event 3D scene scalar covariate vector;

[0049] based on Average effect before triggering

[0050]

[0051] Average effect after triggering

[0052]

[0053] First crossing of zero

[0054]

[0055] in, This indicates the overall average effect before triggering. This indicates the overall average effect after triggering. This indicates the moment when the warning effect first turns into a negative deceleration effect.

[0056] In step 4 above, the function response-function regression model is expressed as follows:

[0057]

[0058] in, This represents the baseline post-response function after controlling the pre-response history. This represents the net post-warning effect function after controlling the pre-response history. The function representing the time-varying influence of baseline velocity on the subsequent response trajectory. This represents the time-varying coefficient vector of the scene covariates on the subsequent response trajectory. This represents a continuous-time variable in the domain before the trigger. Indicates the pre-response cutoff time. Represents the front and back transfer coefficient surfaces. Represents the post-response residual function;

[0059] The transition coefficient surface is expanded using tensor product basis functions.

[0060]

[0061] in, This indicates the number of the basis function in the direction before triggering. This indicates the number of basis functions in the direction before triggering. Indicates the first One pre-trigger direction basis function Indicates the direction basis function number after triggering. Indicates the number of directional basis functions after triggering. Indicates the first One trigger-direction basis function Represents the tensor coefficients.

[0062] In step 4 above, the response curve after constructing a single-event counterfactual event without warning is generated. ,

[0063]

[0064] in, , , and These represent the estimated values ​​of the corresponding parameters;

[0065] Define the net response curve.

[0066]

[0067] in, This indicates the net incremental response resulting from the early warning;

[0068] extract,

[0069]

[0070]

[0071]

[0072]

[0073]

[0074]

[0075] in, Indicates the pre-response intensity. Indicates the net response strength. Indicates response time delay. Indicates peak response. Indicates the effective duration. Indicates the degree of coupling between the front and back ends. Indicates the response threshold. Denotes set measure, This represents the smallest positive number that prevents the denominator from being zero.

[0076] In step 5 above, the index is constructed as a feature vector.

[0077]

[0078] in, Indicates the first The response feature vector of an event;

[0079] The objective function for K-means clustering is,

[0080]

[0081] in, This indicates the total number of driver response types. Indicates the first The set of events corresponding to each driver response type Indicates the first Cluster centers for driver response types; and based on the differences in pre-response intensity, net post-response intensity, response lag, peak response, and effective duration of the cluster centers, driver response types are identified as one or more of the following: advance prediction type, rapid strong response type, delayed correction type, weak response type, and dependent type.

[0082] In step S6 above, according to the first Alarm trigger indicators under individual driver response types Calculate its mean and standard deviation.

[0083]

[0084]

[0085] in, Indicates the first Alarm trigger indicators for each event, Indicates the first A set of events of driver response type Indicates the first The number of events in each driver response type Indicates the first Average alarm and warning values ​​for each driver response type Indicates the first Standard deviation of alarm warnings under individual driver response types;

[0086] Set three different alarm thresholds.

[0087]

[0088]

[0089]

[0090] in, Indicates the first The first sensitive alarm threshold under each driver response type Indicates the first The second standard alarm threshold under each driver response type Indicates the first The third conservative alarm threshold under each driver response type.

[0091] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the time-varying effect quantification method for a driving assistance warning system as described above.

[0092] By adopting the above solution, compared with the prior art, the present invention has the following technical effects:

[0093] (1) This invention expands the evaluation of the effectiveness of driver assistance warning systems from traditional static scalar indicators to a dynamic process over continuous time. By establishing a function response-scalar regression model, it is possible not only to determine whether the warning is effective as a whole, but also to finely characterize when the warning effect begins to appear, how the intensity of the effect changes over time, and how long it can last, which greatly enriches the evaluation dimensions and engineering guidance value of the warning effect.

[0094] (2) By introducing a function response-function regression model and constructing a response curve after the counterfactual event without warning, this invention effectively isolates the behavioral preparation or deceleration trend that the driver has already generated before the warning is triggered. This mechanism can accurately extract the "net incremental response" brought about by the explicit warning, avoiding miscalculation of the driver's own routine preventive behavior as the warning effect, thereby significantly improving the objectivity and accuracy of the assessment of the true causal effect of the warning system.

[0095] (3) This invention accurately identifies driver response types based on extracted net response multidimensional time-varying indicators, and designs three differentiated alarm thresholds for sensitive, standard, and conservative types. This method can adaptively adjust the warning strategy according to different drivers' response capabilities, reaction speeds, and behavioral inertia, effectively reducing false alarm interference while ensuring driving safety, and significantly improving the personalization level and driver acceptance of the driver assistance warning system. Attached Figure Description

[0096] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0097] The technical solution and beneficial effects of the present invention will be described in detail below with reference to the accompanying drawings.

[0098] This invention provides a method for quantifying the time-varying effect of a driving assistance warning system, comprising the following steps:

[0099] S1. Acquire event-level operational data and scene attribute data of the vehicle before and after the triggering of the driver assistance warning system. The event-level operational data includes at least event identifier, timestamp, vehicle speed, vehicle position, and warning issuance marker. The scene attribute data includes at least one or more of time period identifier, road type, traffic status, speed limit information, and road environment attributes. Establish an analysis time window with the warning triggering time as the zero point. By establishing a unified analysis time window, different events can be compared under the same triggering benchmark.

[0100] S2. Perform B-spline smoothing reconstruction on the discrete velocity trajectory after trigger alignment in step S1 to obtain the first... A continuous velocity trajectory function for events, where... Indicate the event number; further construct the normalized velocity change trajectory; and divide the normalized velocity change trajectory into a pre-trigger trajectory and a post-trigger trajectory;

[0101] S3. Based on the normalized velocity change trajectory of the full-time window, establish a function response-scalar regression model to obtain the direct time-varying effect curve of the driving assistance warning system, and extract the average effect before warning triggering, the average effect after warning triggering, and the moment when the warning effect first crosses zero, so as to quantify the overall dynamic impact of the warning system in the entire analysis time window.

[0102] S4. Based on the pre-trigger trajectory and post-trigger trajectory, establish a function response-function regression model to obtain the net post-response effect curve and the transition coefficient surface after controlling the pre-response history; further construct the single-event counterfactual post-response curve without warning. The net response curve is obtained by the difference between the actual triggered trajectory and the response curve after the no-warning counterfactual event.

[0103]

[0104] in, This represents the theoretical post-response trajectory under the same pre-response history and scenario conditions without issuing a warning. This indicates the net incremental response trajectory resulting from the early warning;

[0105] S5. Based on the net response curve, extract the pre-response intensity, net post-response intensity, response lag, peak response, effective duration and pre- and post-coupling degree indicators, and cluster the driver response types based on the indicators to obtain different driver response types.

[0106] S6. Set differentiated alarm thresholds based on the statistical distribution of alarm trigger indicators under different driver response types, and divide alarms into three categories. The alarm thresholds for the three categories are respectively set as the alarm warning mean minus the standard deviation, the alarm warning mean, and the alarm warning mean plus the standard deviation.

[0108] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0109] In one embodiment, obtain the first The observation samples of the event before and after the triggering of the driver assistance warning system, of which the first... The discrete velocity observations of an event within the analysis time window are denoted as follows: ,in, Indicates the event number. Indicates the first The observation point number in each event, This represents the total number of observation points for the event. The time coordinates represent the relative time when the warning was triggered. This represents the corresponding velocity observation value.

[0110] Establish an analysis time window with the warning trigger time as the zero point.

[0111]

[0112] in, Indicates the domain of analysis. This indicates the half-length of the analysis time window. The purpose of this analysis time window is to map different events to the same trigger coordinate system, thereby facilitating the comparison of the velocity changes of different events before and after the warning is triggered.

[0113] Since the original velocity observations may contain uneven sampling, local missing data, and noise interference, B-spline smoothing reconstruction is performed first. Assume the discrete velocity observations satisfy...

[0114]

[0115] in, Indicates the first The event at time The true continuous velocity trajectory function, This represents the observation error term.

[0116] Pick indivual The basis functions of the second-order B-spline form the basis function vector.

[0117]

[0118] in, Indicates the first indivual B-spline basis functions , Indicates transpose. Used to represent a continuous velocity trajectory:

[0119]

[0120] in, Indicates the first A vector of spline coefficients for each event.

[0121] Define velocity observation vector

[0122]

[0123] and spline design matrix By solving

[0124]

[0125] The spline coefficient estimates are obtained, where, express 3D real space, Represents the L2 norm, Indicates the smoothing parameter. This represents the roughness penalty matrix. express 3D real space, Let L2 represent the norm. The first term in this optimization problem ensures that the fitted trajectory matches the observed data, while the second term limits the fluctuation of the fitted trajectory to keep it smooth and stable.

[0126] The corresponding reconstruction speed trajectory is

[0127]

[0128] Preferably, the smoothing parameter Generalized cross-validation was used to determine the balance between fitting accuracy and trajectory smoothness.

[0129]

[0130] in, Represents candidate smoothing parameters The generalized cross-validation value under the following conditions The parameter is Spline coefficient estimation at time, Represents the trace of a matrix;

[0131] in,

[0132]

[0133] This represents the smoothing moment.

[0134] After obtaining the smoothed continuous velocity trajectory, construct the normalized velocity change trajectory.

[0135]

[0136] in, This represents the reference velocity at the left end of the analysis time window. Indicates the first The trajectory of velocity change of each event relative to a baseline velocity. This normalized trajectory eliminates differences in initial velocity levels between different events, allowing subsequent models to focus primarily on the changes before and after the warning trigger. The trajectory before triggering is further defined.

[0137]

[0138] and the trajectory after triggering

[0139]

[0140] in, It indicates the speed change trajectory before the warning is triggered, and is used to depict whether the driver has already made behavioral preparations before the warning is triggered; It represents the speed change trajectory after the warning is triggered, and is used to depict the driver's response process after the warning is triggered.

[0141] Based on this, a function response-scalar regression model is established to quantify the direct time-varying effect of the driving assistance warning system throughout the entire analysis time window.

[0142]

[0143] in, This represents the reference response function under reference conditions. This represents the direct time-varying effect function of the warning issuance on the normalized velocity change trajectory. The time-varying influence function representing the baseline velocity, The time-varying coefficient vector representing the scene covariate vector. Represents the residual function; This indicates that a warning has been issued. This indicates that a warning has been issued. This indicates that no warning was issued or a silent comparison was performed; Indicates the first The baseline velocity of each event, Indicates the first an event A 3D scene covariate vector. Obtained through estimation. The curve can depict the continuous effect of the early warning throughout the entire time window.

[0144] To further quantify the direct time-varying effect curve, by The following summary indicators can be constructed:

[0145]

[0146]

[0147]

[0148] in, This indicates the overall average effect before the warning is triggered. This indicates the overall average effect after the warning is triggered. This indicates the moment when the warning effect first turns into a negative deceleration effect.

[0149] To separate pre-existing behavioral trends before the warning is triggered, a function response-function regression model is further established.

[0150]

[0151] in, This represents the baseline post-response function after controlling the pre-response history. This represents the net post-warning effect function after controlling the pre-response history. The function representing the time-varying influence of baseline velocity on the subsequent response trajectory. This represents the time-varying coefficient vector of the scene covariates on the subsequent response trajectory. This represents a continuous-time variable in the domain before the trigger. Indicates the pre-response cutoff time. Represents the front and back transfer coefficient surfaces. This represents the post-response residual function. The model is used to separate the behavioral history before the warning is triggered from the behavioral changes after the warning is triggered.

[0152] The transition coefficient surface is preferably expanded using tensor product basis functions.

[0153]

[0154] in, This indicates the number of the basis function in the direction before triggering. This indicates the number of basis functions in the direction before triggering. Indicates the first One pre-trigger direction basis function Indicates the direction basis function number after triggering. Indicates the number of directional basis functions after triggering. Indicates the first One trigger-direction basis function Represents the tensor coefficients. Further define the projection characteristics.

[0155]

[0156] in, Indicates the first The event in the 1st The projection of the basis function in the direction of the triggering.

[0157] Construct a counterfactual response curve without warning based on the model.

[0158]

[0159] in, , , and These represent the estimated values ​​of the corresponding parameters.

[0160] Further, the net response curve was obtained.

[0161]

[0162] in, This indicates the net incremental response brought about by the early warning after excluding the influence of pre-response history and scenarios.

[0163] The following indicators were extracted based on the net response curve:

[0164]

[0165]

[0166]

[0167]

[0168]

[0169]

[0170] in, Indicates the pre-response intensity. Indicates the net response strength. Indicates response time delay. Indicates peak response. Indicates the effective duration. Indicates the degree of coupling between the front and back ends. Indicates the response threshold. Denotes set measure, This represents the smallest positive number that prevents the denominator from being zero.

[0171] The above indicators are constructed as response feature vectors.

[0172]

[0173] in, Indicates the first The response feature vectors of each event are analyzed; and a clustering algorithm is used to cluster these response feature vectors to identify the driver's response type. Preferably, when K-means clustering is used, its objective function is...

[0174]

[0175] in, Indicates the total number of driver response types. Indicates the first The set of events corresponding to each driver response type Indicates the first Cluster centers for driver response types;

[0176] Based on the differences in pre-response intensity, net post-response intensity, response lag, peak response, and effective duration of cluster centers, driver response types can be identified as one or more of the following: anticipatory, rapid strong response, delayed correction, weak response, and dependent.

[0177] For the Alarm trigger indicators under individual driver response types Calculate its mean and standard deviation:

[0178]

[0179]

[0180] in, Indicates the first Alarm trigger indicators for each event, Indicates the first A set of events of driver response type Indicates the first The number of events in each driver response type Indicates the first Average alarm and warning values ​​for each driver response type Indicates the first Standard deviation of alarm warnings under individual driver response types;

[0181] Further construct three types of differentiated early warning thresholds:

[0182]

[0183]

[0184]

[0185] in, Indicates the first The first sensitive alarm threshold under each driver response type Indicates the first The second standard alarm threshold under each driver response type Indicates the first The third type of conservative warning threshold is set for each driver response type. By setting different thresholds for different driver response types, the driver assistance warning system can be configured in a differentiated manner.

[0186] Those skilled in the art should understand that any equivalent substitutions made to the B-spline smoothing method, the function regression model form, the response feature construction method, and the threshold setting method without departing from the concept of the present invention should fall within the protection scope of the present invention.

[0187] This invention also provides another computer device, including a processor and a memory configured to store a computer program capable of running on the processor; wherein, when the processor is configured to run the computer program, it performs the method steps described in the foregoing embodiments.

[0188] In an exemplary embodiment, the present invention also provides a computer-readable storage medium for storing a computer program.

[0189] Optionally, the computer-readable storage medium can be applied to any of the methods in the embodiments of the present invention, and the computer program causes the computer to execute the corresponding processes implemented by the processor in the various methods of the embodiments of the present invention. For the sake of brevity, these will not be described in detail here.

[0190] In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

Claims

1. A method for quantifying the time-varying effect of a driver assistance warning system, characterized in that... Includes the following steps: Step 1: Obtain event-level operational data and scene attribute data of the vehicle before and after the driver assistance warning system is triggered, and establish an analysis time window with the warning trigger time as the zero point; Step 2: Based on the data within the analysis time window, construct the normalized velocity change trajectory and divide it into the pre-trigger trajectory and the post-trigger trajectory; Step 3: Establish a function response-scalar regression model based on the normalized velocity change trajectory of the full time window to obtain the direct time-varying effect curve of the driving assistance warning system, and extract the average effect before warning triggering, the average effect after warning triggering, and the moment when the warning effect first crosses zero. Step 4: Based on the pre-trigger trajectory and the post-trigger trajectory, establish a function response-function regression model to obtain the net post-response effect curve and the front-to-back transition coefficient surface after controlling the pre-response history; further construct a single-event no-warning counterfactual post-response curve, and obtain the net response curve from the difference between the actual post-trigger trajectory and the no-warning counterfactual post-response curve; Step 5: Extract the pre-response intensity, net post-response intensity, response lag, peak response, effective duration and pre- and post-coupling degree indicators based on the net response curve, and cluster the driver response types based on the indicators to obtain different driver response types. Step 6: Set differentiated alarm thresholds based on the statistical distribution of alarm triggering indicators under different driver response types, and divide alarms into three categories. The alarm thresholds for the three categories are respectively set as the alarm warning mean minus the standard deviation, the alarm warning mean, and the alarm warning mean plus the standard deviation.

2. The method as described in claim 1, characterized in that: In step 1, the event-level operational data includes event identifier, timestamp, vehicle speed, vehicle location, and warning issuance marker, while the scene attribute data includes at least one of time period identifier, road type, traffic status, speed limit information, and road environment attributes.

3. The method as described in claim 1, characterized in that: The specific content of step 2 is as follows: Step 21: Based on the data within the analysis time window, perform B-spline smoothing reconstruction on the triggered aligned discrete velocity trajectory to obtain the first... Continuous velocity trajectory function of an event ,in, Indicates the event number. A continuous time variable representing the relative warning trigger time. Indicates the first The event at time Continuous velocity; Step 22, construct the normalized velocity change trajectory. ， in, Indicates the first The baseline velocity of an event at the left end of the analysis time window. Indicates the first The velocity change trajectory of an event relative to a reference velocity; This indicates the half-length of the analysis time window; Step 23: Divide the normalized velocity change trajectory into a pre-trigger trajectory and a post-trigger trajectory. ， ， in, This indicates the trajectory of speed changes before the warning is triggered. This indicates the trajectory of speed changes after the warning is triggered.

4. The method as described in claim 3, characterized in that: In step 21, let the first... The discrete velocity observations of each event within the analysis time window are as follows: ,in, Indicates the first The observation point number in each event, Indicates the first The total number of observation points for each event. Indicates the first Time coordinates of each observation point Indicates the first Velocity observation values ​​at each observation point; Assume that discrete observations satisfy, ， in, Represents the observation error term; take indivual The basis functions of the second-order B-spline form the basis function vector. ， in, Indicates the first indivual B-spline basis functions , Indicates transpose; The continuous velocity trajectory function is expressed as follows: ， in, Indicates the first The spline coefficient vector of events; let ， in, Represents the velocity observation vector. Represents the spline design matrix; Solve ， The spline coefficient estimates are obtained, where, express 3D real space, Represents the L2 norm, Indicates the first Smoothing parameters for each event, Represents the roughness penalty matrix; The corresponding reconstruction speed trajectory is ， Smoothing parameters Determined through generalized cross-validation ， in, Represents candidate smoothing parameters The generalized cross-validation value under the following conditions The parameter is Spline coefficient estimation at time, Represents the trace of a matrix. ， This represents a smoothing matrix.

5. The method as described in claim 1, characterized in that: In step 3, the function response-scalar regression model is expressed as follows: ， in, This represents the reference response function under reference conditions. This represents the direct time-varying effect function of the warning issuance on the normalized velocity change trajectory. The time-varying influence function representing the baseline velocity, The time-varying coefficient vector representing the scene covariate vector. Represents the residual function; This indicates that a warning has been issued. This indicates that a warning has been issued. This indicates that no warning was issued or a silent comparison was performed; Indicates the first The baseline velocity of each event, Indicates the first an event 3D scene scalar covariate vector; based on Average effect before triggering ， Average effect after triggering ， First crossing of zero ， in, This indicates the overall average effect before triggering. This indicates the overall average effect after triggering. This indicates the moment when the warning effect first turns into a negative deceleration effect.

6. The method as described in claim 1, characterized in that: In step 4, the function response-function regression model is expressed as follows: ， in, This represents the baseline post-response function after controlling the pre-response history. This represents the net post-warning effect function after controlling the pre-response history. The function representing the time-varying influence of baseline velocity on the subsequent response trajectory. This represents the time-varying coefficient vector of the scene covariates on the subsequent response trajectory. This represents a continuous-time variable in the domain before the trigger. Indicates the pre-response cutoff time. Represents the front and back transfer coefficient surfaces. Represents the post-response residual function; The transition coefficient surface is expanded using tensor product basis functions. ， in, This indicates the number of the basis function in the direction before triggering. Indicates the number of basis functions in the direction before triggering. Indicates the first One pre-triggering direction basis function Indicates the direction basis function number after triggering. Indicates the number of directional basis functions after triggering. Indicates the first One trigger-direction basis function Represents the tensor coefficients.

7. The method as described in claim 1, characterized in that: In step 4, a single-event counterfactual response curve without warning is constructed. , ， in, , , and These represent the estimated values ​​of the corresponding parameters; Define the net response curve. ， in, This indicates the net incremental response resulting from the early warning; extract, ， ， ， ， ， ， in, Indicates the pre-response intensity. Indicates the net response strength. Indicates response time delay. Indicates peak response. Indicates the effective duration. Indicates the degree of coupling between the front and back ends. Indicates the response threshold. Denotes set measure, This represents the smallest positive number that prevents the denominator from being zero.

8. The method as described in claim 1, characterized in that: In step 5, the index is constructed as a feature vector. ， in, Indicates the first The response feature vector of an event; The objective function for K-means clustering is, ， in, Indicates the total number of driver response types. Indicates the first The set of events corresponding to each driver response type Indicates the first Cluster centers for driver response types; and based on the differences in pre-response intensity, net post-response intensity, response lag, peak response, and effective duration of the cluster centers, driver response types are identified as one or more of the following: advance prediction type, rapid strong response type, delayed correction type, weak response type, and dependent type.

9. The method as described in claim 1, characterized in that: In step S6, according to the first Alarm trigger indicators under individual driver response types Calculate its mean and standard deviation. ， ， in, Indicates the first Alarm trigger indicators for each event, Indicates the first A set of events of driver response type Indicates the first The number of events in each driver response type Indicates the first Average alarm and warning values ​​for each driver response type Indicates the first Standard deviation of alarm warnings under individual driver response types; Set three different alarm thresholds. ， ， ， in, Indicates the first The first sensitive alarm threshold under each driver response type Indicates the first The second standard alarm threshold under each driver response type Indicates the first The third conservative alarm threshold under each driver response type.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the method for quantifying the time-varying effects of a driving assistance warning system as described in any one of claims 1 to 9.

Images