A risk assessment method and system for autonomous driving of intelligent connected vehicles

Abstract

The present invention proposes a method and system for risk assessment of autonomous driving of intelligent networked vehicles, wherein the method includes: obtaining the basic configuration and historical driving data of the target vehicle; inputting the basic configuration into an autonomous driving safety performance scoring model pre-built based on a variety of road scenarios to obtain a comprehensive safety performance score of the target vehicle; inputting the comprehensive safety performance score and historical driving data into a dynamic disengagement model pre-built based on driving behavior to obtain the disengagement rate of autonomous driving; inputting the disengagement rate into an accident rate conversion model pre-built based on adjustment rules to obtain the accident rate of the target vehicle. The present invention first performs static evaluation in multiple scenarios based on the configuration information of the intelligent networked vehicle, and then further combines the static evaluation results and driving data to realize dynamic evaluation of the target vehicle and obtain the dynamic disengagement rate, and finally predicts its accident rate based on the dynamic disengagement rate as the main basis to achieve a comprehensive and relatively accurate risk assessment.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for assessing the risk of autonomous driving of an intelligent networked vehicle. Background Art

[0002] In recent years, intelligent connected vehicles have developed rapidly in my country. In 2023, the market penetration rate of intelligent connected passenger vehicles equipped with assisted autonomous driving systems in my country has exceeded 50%, and it is expected to grow to 70% in 2025. Many local intelligent connected vehicle access testing policies mention the demand for insurance, and encourage the insurance industry to develop exclusive insurance products for intelligent connected vehicles. At present, there are no exclusive insurance products for intelligent connected vehicles. Most high-end intelligent connected vehicles are insured with ordinary vehicle insurance during testing. The risk assessment system for ordinary vehicle insurance is mature. It is based on the actuarial pricing model method of generalized linear models, uses a large amount of historical policy data and claims data to model, and evaluates the coefficients of risk factors. Common risk factors include age and gender, vehicle price, vehicle model, number of seats, vehicle age, etc. However, ordinary vehicle insurance products are not actually suitable for intelligent connected vehicles for the following reasons:

[0003] 1. There is a lack of data related to intelligent connected vehicles, and the traditional risk assessment system cannot be used;

[0004] 2. The coverage of ordinary vehicle insurance products is not comprehensive enough, and insurance products are unable to conduct risk assessments on the safety performance of autonomous driving, resulting in inaccurate pricing of existing vehicle insurance products for smart connected vehicles. Summary of the invention

[0005] In order to achieve a comprehensive analysis of the safety performance of autonomous driving to achieve a comprehensive and relatively accurate risk assessment, in a first aspect of the present invention, a method for risk assessment of autonomous driving of an intelligent connected vehicle is proposed, the method comprising: obtaining basic configuration and historical driving data of a target vehicle; inputting the basic configuration into an autonomous driving safety performance scoring model pre-constructed based on a variety of road scenarios to obtain a comprehensive safety performance score of the target vehicle; inputting the comprehensive safety performance score and the historical driving data into a dynamic disengagement model pre-constructed based on driving behavior to obtain a disengagement rate of autonomous driving; and inputting the disengagement rate into an accident rate conversion model pre-established based on adjustment rules to obtain the accident rate of the target vehicle.

[0006] In one or more embodiments, the method also includes constructing an autonomous driving safety performance scoring model through the following steps, including: dividing the safety performance of autonomous driving into multiple first-level scoring indicators according to the autonomous driving level of the vehicle; dividing each of the first-level scoring indicators into multiple second-level scoring indicators according to multiple road scenarios; determining the basic configuration or basic configuration combination related to the second-level scoring indicator, wherein the basic configuration combination includes basic hardware configuration and / or basic software configuration; testing the second-level safety score of the basic configuration combination under the second-level scoring indicator, and associating and storing the second-level safety score with the basic configuration combination; designing a second-level weight factor according to the basic hardware configuration level, quantity and / or whether there is a basic software configuration corresponding to the second-level scoring indicator in the basic configuration combination; designing a first-level weight for each of the first-level scoring indicators according to the importance of the first-level scoring indicator to the safety performance of the vehicle's autonomous driving; and constructing the autonomous driving safety performance scoring model based on the first-level weight, the second-level weight factor, the second-level safety score and the corresponding basic configuration combination.

[0007] In one or more embodiments, the safety performance of autonomous driving is divided into multiple first-level scoring indicators according to the autonomous driving level of the vehicle, including: dividing the safety performance of autonomous driving into following capability, combined control capability, emergency avoidance capability, and monitoring interaction capability.

[0008] In one or more embodiments, each of the first-level scoring indicators is divided into multiple second-level scoring indicators according to multiple road scenes, including: dividing the following vehicle capability into the front vehicle being stationary, the front vehicle being at low speed, the front vehicle being decelerated, the front vehicle cutting in, the front vehicle cutting out and / or following the front vehicle to start and stop; dividing the combined control capability into lane centering control, traffic congestion assistance, high-speed driving assistance, non-interference vehicle lane changing and / or interference vehicle lane changing; dividing the emergency avoidance capability into obstructed pedestrian crossing, pedestrian crossing at night, two-wheeled vehicle crossing, accident vehicle identification and response and / or cone barrel identification and response; dividing the monitoring interaction capability into driver eyes closed monitoring, driver head down monitoring, driver status monitoring interaction and / or alarm event monitoring interaction.

[0009] In one or more embodiments, the basic configuration is input into an autonomous driving safety performance scoring model pre-constructed based on multiple road scenarios to obtain a comprehensive safety performance score of the target vehicle, including: extracting keywords as classification labels for the basic configuration based on the functional description of the basic configuration; clustering the classification labels based on the primary scoring indicators and the secondary scoring indicators and obtaining the basic configurations corresponding to the classification labels under the same cluster to form a basic configuration combination; determining the basic hardware configuration level, quantity and / or whether there is a basic software configuration corresponding to the secondary scoring indicator in the basic configuration and comparing it with the basic configuration combination information stored in the autonomous driving safety performance scoring model to determine the secondary weight; determining the primary weight based on the primary scoring indicator corresponding to the configuration combination; calculating the safety performance score of the target vehicle in each road scenario based on the primary weight, the secondary weight and the secondary safety score; and calculating the average safety performance score of the target vehicle in all road scenarios to obtain the comprehensive safety performance score.

[0010] In one or more embodiments, the method also includes constructing a dynamic disengagement model through the following steps, including: obtaining historical driving data of the target vehicle, analyzing and extracting key driving behaviors that affect the safety performance of autonomous driving in the historical driving data; during the test, recording the first time point of the critical disengagement; marking the key driving behaviors extracted within a preset time before and after the first time point, and using them as training samples; training an autonomous driving critical disengagement classification model based on the training samples to obtain an initial dynamic disengagement model; and configuring different classification influence weights for the initial dynamic disengagement model according to a preset scoring range to construct the dynamic disengagement model.

[0011] In one or more embodiments, the comprehensive safety performance score and the historical driving data are input into a dynamic disengagement model pre-constructed based on driving behavior to obtain the disengagement rate of autonomous driving, including: determining the classification influence weight of the dynamic disengagement model according to a preset score range of the comprehensive safety performance score to determine the dynamic disengagement model used this time; obtaining the historical driving data of the target vehicle and extracting key driving behaviors; inputting the key driving behaviors into the dynamic disengagement model to determine the disengagement points in the key driving behaviors; and counting the mileage at different average speeds, and obtaining the disengagement rate of autonomous driving by comparing the number of disengagement points occurring in the corresponding mileage with the corresponding mileage.

[0012] In one or more embodiments, acquiring the historical driving data of the target vehicle and extracting key driving behaviors include: extracting the average speed of the target vehicle exceeding a specified speed threshold, the average acceleration exceeding a preset acceleration threshold, sudden braking and / or sharp turns.

[0013] In one or more embodiments, the method also includes constructing an accident rate conversion model through the following steps, including: collecting the disengagement rate data of the insured vehicles before the accident, and analyzing the disengagement rate-accident rate conversion curve at different average speeds; configuring an increaseable and modifiable adjustment factor for the disengagement rate-accident rate conversion curve, each adjustment factor corresponding to an adjustment rule; configuring a callback function for each adjustment factor to adjust the adjustment factor when the adjustment rule changes; wherein the factors involved in the adjustment rule include: the number of accidents and the total compensation amount in the previous year, the city where the vehicle is located and / or the commonly used driving routes.

[0014] In a second aspect of the present invention, a risk assessment system for autonomous driving of intelligent connected vehicles is proposed, comprising: a data collection module, configured to obtain basic configuration and historical driving data of a target vehicle; a safety performance scoring module, configured to input the basic configuration into an autonomous driving safety performance scoring model pre-constructed based on a variety of road scenarios to obtain a comprehensive safety performance score of the target vehicle; a disengagement rate prediction module, configured to input the safety performance score and the historical driving data into a dynamic disengagement model pre-constructed based on driving behavior to obtain the disengagement rate of autonomous driving; and an accident rate prediction module, configured to input the disengagement rate into an accident rate conversion model pre-established based on adjustment rules to obtain the accident rate of the target vehicle.

[0015] The beneficial effects of the present invention include: the present invention first performs static evaluation in multiple scenarios based on the configuration information of the intelligent connected vehicle, and then on this basis, further combines the static evaluation results and driving data to realize dynamic evaluation of the target vehicle and predict the dynamic disengagement rate of the target vehicle, and finally uses the dynamic disengagement rate as the main basis for analyzing its accident rate to realize a comprehensive and relatively accurate risk assessment of the target vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is a workflow diagram of an embodiment of a method for risk assessment of autonomous driving of an intelligent networked vehicle of the present invention;

[0018] Figure 2 It is a schematic diagram of the structure of the three-layer intelligent networked vehicle risk assessment model of the present invention;

[0019] Figure 3 This is a schematic diagram of the structure of an embodiment of a risk assessment system for autonomous driving of an intelligent connected vehicle according to the present invention. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the embodiments of the present invention are further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0021] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are for distinguishing two non-identical entities with the same name or non-identical parameters. It can be seen that "first" and "second" are only for the convenience of expression and should not be understood as limitations on the embodiments of the present invention. The subsequent embodiments will not explain this one by one.

[0022] In order to comprehensively evaluate the risk of connected vehicles, the present invention proposes a three-layer intelligent connected vehicle risk assessment model. The first layer and the second layer are used to evaluate the static configuration and dynamic driving ability of the vehicle, respectively. The first layer considers the static comprehensive score of the intelligent connected vehicle configuration to adapt to various road scenes, and transmits it to the second layer as a risk factor to determine the model used in the second layer. The second layer is used to determine whether the monitored disengagement point is a non-critical disengagement or a critical disengagement according to the extracted key driving behavior, and record the number of disengagements and calculate the disengagement rate when it is judged to be a critical disengagement for evaluating the dynamic safety performance of the target vehicle. The third layer is used to determine the accident rate according to the preset disengagement rate-accident rate conversion curve and the corresponding adjustment rules, and then perform risk assessment based on the accident rate. In this way, the present invention realizes a comprehensive evaluation of the driving system, and predicts the accident rate by analyzing the disengagement rate of the target vehicle in the year to overcome the problem that the risk assessment cannot be performed due to the small amount of accident data of autonomous driving. It should be noted that in the later process of using the model for evaluation, the disengagement rate is only a counting and expression method used to predict the risk in the present invention, and does not represent the actual situation of the autonomous driving system being out of control. For a more specific understanding, please refer to the following implementation plan:

[0023] For more specific implementation methods, please refer to the following embodiments:

[0024] In one embodiment, see Figure 1The workflow of a risk assessment method for autonomous driving of an intelligent connected vehicle of the present invention includes: step S1, obtaining the basic configuration and historical driving data of the target vehicle; step S2, inputting the basic configuration into an autonomous driving safety performance scoring model pre-built based on multiple road scenarios to obtain a comprehensive safety performance score of the target vehicle; step S3, inputting the comprehensive safety performance score and historical driving data into a dynamic disengagement model pre-built based on driving behavior to obtain the disengagement rate of autonomous driving; and step S4, inputting the disengagement rate into an accident rate conversion model pre-built based on adjustment rules to obtain the accident rate of the target vehicle.

[0025] As described above, the model in this implementation method fully considers the static configuration and dynamic control capabilities of the intelligent connected vehicle (hereinafter referred to as the target vehicle), thereby achieving a comprehensive assessment of the vehicle's safety, and predicting the vehicle's disengagement rate based on the assessment results and then converting it into an accident rate, thereby achieving a more accurate assessment of the accident rate.

[0026] In one embodiment, the method of the present invention further comprises constructing a performance scoring model for the autonomous driving vehicle by the following steps, including:

[0027] Step 100, dividing the safety performance of the autonomous driving into a plurality of first-level scoring indicators according to the autonomous driving level of the vehicle; specifically, since there are multiple levels of autonomous driving, in order to achieve a comprehensive evaluation of vehicles of different levels, the present invention proposes dividing the first-level scoring indicators according to the autonomous driving level. Taking the L2 level autonomous driving adopted by the current mainstream networked vehicles as an example, it is still a kind of assisted driving and cannot be completely separated from the driver's attention. According to its ability to achieve basic functions of assisted driving, such as following vehicle capability, combined control capability, emergency avoidance capability and driver monitoring interaction capability, it is divided into the above four first-level scoring indicators, and then each first-level indicator is refined into a plurality of second-level indicators.

[0028] In an optional embodiment, the classification of the first / second / third level scoring indicators for vehicles of different autonomous driving levels is shown in the following table:

[0029]

[0030]

[0031]

[0032]

[0033] In the above table, each test result will be saved as a scoring element in the first-level model - the performance scoring model, and each test is actually aimed at different software and hardware configuration combinations or the same configuration combination in different aspects, that is, each first-level scoring indicator corresponds to a set of software and hardware configuration combinations. For example, different road scenarios are considered in each second-level / third-level scoring in the above table. The purpose of the above test is to test the safety performance of a certain software and hardware configuration combination in different road scenarios, in order to subsequently realize a static evaluation of the overall safety of the target vehicle (i.e., the vehicle to be evaluated) based on the software and hardware configuration of the target vehicle.

[0034] Step 110: Divide each first-level scoring indicator into multiple second-level scoring indicators according to multiple road scenarios; specifically, the second-level scoring indicators are used to further subdivide the scenarios adapted by each configuration or configuration combination based on the software and hardware configuration combination determined by the first-level indicators in order to enable a more detailed evaluation of the target vehicle.

[0035] Step 120: determine the basic configuration combination related to the secondary scoring index, wherein the basic configuration combination includes the basic hardware configuration and / or the basic software configuration. Specifically, the implementation of step 120 requires obtaining the driving data of the modeling test vehicle, first analyzing and determining the basic configuration combination related to each primary scoring index. For example, in the process of following the vehicle capability test, analyzing which type of related data is generated in the largest amount, analyzing which software and hardware configurations are the most active, etc.; then, further analyzing the sub-category basic configuration combination related to each secondary scoring index.

[0036] Step 130: Test the secondary safety score of the basic configuration combination under the secondary scoring indicator, and store the secondary safety score in association with the basic configuration combination; if applicable, the test may be conducted in a virtual environment or on a closed road. This implementation further subdivides the road scene corresponding to the secondary scoring indicator, and needs to test the score of the third-level scoring indicator corresponding to each secondary scoring indicator, and assign different weights to each third-level scoring indicator, and then determine the secondary safety score by calculating the weighted average.

[0037] Step 140, design a secondary weight factor according to the basic hardware configuration level, quantity and / or whether there is a basic software configuration corresponding to the secondary scoring index in the basic configuration combination; specifically, in order to cope with the configuration differences between different intelligent connected vehicles and subsequent configuration upgrades, the present invention makes adaptive adjustments by designing some secondary weight factors. In addition, since the present application has made a fine division of the configuration combination through multi-level scoring indicators in the process of building the model, when the configuration of the target vehicle is different from the configuration of the test vehicle used to build the model, the difference under a configuration combination corresponding to the secondary scoring index will not be too large, and the difference between the two can be adjusted by the secondary weight factor.

[0038] In an optional embodiment, taking the vehicle following capability as an example, the vehicle following capability mainly depends on the hardware configuration of the laser radar and the related software configuration. In the model building stage, the level, quantity and whether there is a basic software configuration corresponding to the secondary scoring indicator of the laser radar of the test vehicle are saved; after obtaining the basic configuration of the vehicle to be evaluated, if the laser radar level in the configuration is higher, the quantity is greater, and there is a corresponding basic software configuration, then the secondary weight with a configuration value greater than 1 is set, otherwise, the secondary weight with a configuration value less than or equal to 1 is set. For example, let C1 be the weight corresponding to the laser radar level, C2 be the weight corresponding to the number of laser radars, and C3 be the weight corresponding to whether there is a corresponding basic software configuration; accordingly, the function expression of the comprehensive score of the sub-scenario corresponding to each primary scoring indicator in the autonomous driving safety performance scoring model is:

[0039]

[0040]

[0041] Among them, Pk is the first-level safety score, x is the basic configuration of the target vehicle, and C is the second-level weight.

[0042] Step 150: Design a first-level weight for each first-level scoring indicator according to the importance of the first-level scoring indicator to the safety performance of the vehicle's autonomous driving; specifically, the importance of the first-level indicator can be evaluated by reference to its common usage or the severity of the consequences caused by a failure of the first-level scoring indicator.

[0043] Step 160: Construct an autonomous driving safety performance scoring model based on the primary weight, the secondary weight factor, the secondary safety score, and the corresponding basic configuration or basic configuration combination. Specifically, each secondary safety score and the corresponding sub-class basic configuration combination can be regarded as a matching sub-model (i.e., a sub-scenario), which is used to determine the corresponding secondary weight by judging the matching degree between the configuration of the vehicle to be evaluated and the configuration or configuration combination in the matching sub-model, thereby estimating the static score of the vehicle in the road sub-scenario.

[0044] In one embodiment, the safety performance of autonomous driving is divided into multiple first-level scoring indicators according to the autonomous driving level of the vehicle, including: dividing the safety performance of autonomous driving into following capability, combined control capability, emergency avoidance capability and monitoring interaction capability; dividing each first-level scoring indicator into multiple second-level scoring indicators according to multiple road scenarios, including: dividing the following capability into the front vehicle being stationary, the front vehicle being at low speed, the front vehicle being decelerated, the front vehicle cutting in, the front vehicle cutting out and / or following the front vehicle start and stop; dividing the combined control capability into lane centering control, traffic congestion assistance, high-speed driving assistance, lane changing without interference and / or lane changing with interference; dividing the emergency avoidance capability into obstructed pedestrian crossing, pedestrian crossing at night, two-wheeled vehicle crossing, accident vehicle recognition and response and / or cone barrel recognition and response; dividing the monitoring interaction capability into driver eyes closed monitoring, driver head down monitoring, driver status monitoring interaction and / or alarm event monitoring interaction.

[0045] In one embodiment, the basic configuration is input into an autonomous driving safety performance scoring model pre-constructed based on a variety of road scenarios to obtain a comprehensive safety performance score of the target vehicle, including: step S2.1, extracting keywords as classification labels of the basic configuration according to the functional description of the basic configuration; step S2.2, clustering the classification labels according to the primary scoring indicators and the secondary scoring indicators and obtaining the basic configurations of the corresponding classification labels under the same cluster to form a basic configuration combination; step S2.3, determining the basic hardware configuration level, quantity and / or whether there is a basic software configuration corresponding to the secondary scoring indicator in the basic configuration and comparing it with the basic configuration combination information stored in the autonomous driving safety performance scoring model to determine the secondary weight; step S2.4, determining the primary weight according to the primary scoring indicator corresponding to the configuration combination; step S2.5, calculating the safety performance score of the target vehicle in each road scenario according to the primary weight, the secondary weight and the secondary safety score; step S2.6, calculating the average safety performance score of the target vehicle in all road scenarios to obtain a comprehensive safety performance score.

[0046] Specifically, in the process of executing step S2.3, it is also necessary to compare the basic configuration combination clustered according to the secondary scoring index with the sub-class basic configuration combination corresponding to the secondary safety score in the sub-scenario element of the model. If the corresponding basic configuration is missing, the secondary weight is lowered accordingly. If there is a redundant configuration, the secondary weight is increased accordingly.

[0047] The calculation formula of step S2.6 is:

[0048]

[0049] in, is the first-level weight, is the corresponding first-level safety score.

[0050] In one embodiment, the method of the present invention also includes constructing a dynamic disengagement model through the following steps, including: step 200, obtaining historical driving data of the target vehicle, analyzing and extracting key driving behaviors that affect the safety performance of autonomous driving in the historical driving data; step 210, recording the first time point of the key disengagement during the test; step 220, marking the key driving behaviors extracted within a preset time before and after the first time point, and using them as training samples; step 230, training an autonomous driving key disengagement classification model based on the training samples to obtain an initial dynamic disengagement model; step 240, configuring different classification influence weights for the initial dynamic disengagement model according to a preset scoring range to construct a dynamic disengagement model.

[0051] Specifically, the disengagement rate refers to the state of disengagement from the autonomous driving state, which usually means that the safety officer subjectively judges that the vehicle is about to collide and thus takes over the driving rights of the vehicle. Disengagement is divided into critical disengagement and non-critical disengagement. For example, if the safety officer needs to stop and rest, he will also take over the driving rights of the vehicle. This situation is called non-critical disengagement; secondly, the more critical disengagements per unit distance, the more imperfect the autonomous driving algorithm of the car is, and the higher the risk. Due to the huge amount of original disengagement data, there are a large number of non-critical disengagement scenarios. These risk-free disengagement scenarios will distort the analysis results of the model and need to be eliminated through the dynamic disengagement model.

[0052] In addition, different preset scoring ranges are correspondingly set with different classification influence weights, and the classification influence weights will affect the judgment of the nature of the disengagement point under the same conditions. Among them, the setting logic of the corresponding relationship between the classification influence weight and the scoring range is that the higher the comprehensive safety performance score of the target vehicle, the lower the probability of judging a specific key driving behavior as a disengagement point. It should be pointed out that the same driving behavior will cause different disengagement risks if it appears in different orders. The present invention will train such scenarios during the training process of the dynamic disengagement model.

[0053] In one embodiment, the comprehensive safety performance score and historical driving data are input into a dynamic disengagement model pre-constructed based on driving behavior to obtain the disengagement rate of autonomous driving, including: step S3.1, determining the classification influence weight of the dynamic disengagement model according to the preset score range of the comprehensive safety performance score to determine the dynamic disengagement model used this time; step S3.2, obtaining the historical driving data of the target vehicle and extracting key driving behaviors; step S3.3, inputting the key driving behaviors into the dynamic disengagement model to determine the disengagement points in the key driving behaviors (i.e., there may be situations where the driver needs to intervene between the key driving behaviors); step S3.4, counting the mileage at different average speeds, and obtaining the disengagement rate of autonomous driving by dividing the number of disengagement points in the corresponding mileage by the corresponding mileage.

[0054] In one embodiment, historical driving data of the target vehicle is obtained to extract key driving behaviors, including: extracting the average speed of the target vehicle exceeding a specified speed threshold, the average acceleration exceeding a preset acceleration threshold, sudden braking and / or sharp turns. Specifically, the determination of the disengagement rate may also refer to other factors, such as the mileage of the target vehicle, maintenance records, etc.

[0055] In one embodiment, the method of the present invention also includes constructing an accident rate conversion model through the following steps, including: collecting the disengagement rate data of the insured vehicles before the accident, and analyzing the disengagement rate-accident rate conversion curve at different average speeds; configuring an increaseable and modifiable adjustment factor for the disengagement rate-accident rate conversion curve, each adjustment factor corresponding to an adjustment rule; configuring a callback function for each adjustment factor to adjust the adjustment factor when the adjustment rule changes; wherein the factors involved in the adjustment rule include: the number of accidents and the total compensation amount in the previous year, the city where the vehicle is located and / or the commonly used driving routes.

[0056] In some embodiments, the adjustment rules include: if the number of accidents and the total compensation amount in the previous year show an upward trend, then the adjustment factor will be appropriately increased to increase the proportion of converting the escape rate into the accident rate, otherwise the proportion will be reduced; the adjustment factor is set according to the proportion of national highways and expressways in commonly used driving routes. The higher the proportion of national highways, the larger the adjustment factor will be appropriately increased to increase the proportion of converting the escape rate into the accident rate, otherwise the proportion will be reduced.

[0057] In a further embodiment, the disengagement rate is input into an accident rate conversion model pre-established based on adjustment rules to obtain the accident rate of the target vehicle, including: dividing the mileage of each day into multiple segments, calculating the average driving speed of each mileage segment and counting the disengagement rate; obtaining a disengagement rate-accident rate conversion curve corresponding to the above average driving speed, and determining the influencing factor of the disengagement rate-accident rate conversion curve according to the number of accidents and the total compensation amount in the previous year, the city where the vehicle is located and / or the commonly used driving routes, and dynamically calculating the accident rate of the day based on the disengagement rate, the disengagement rate-accident rate conversion curve and the influencing factor occurring on the day; finally, the annual accident rate is obtained by accumulating and averaging the accident rates of each day for risk assessment.

[0058] The structure of the three-layer intelligent connected vehicle risk assessment model of the present invention is shown as follows: Figure 2 As shown in the figure, Dn in the dynamic disengagement model represents the dynamic disengagement rate, Er represents multiple driving behaviors, and x represents the influence weights between driving behaviors; Cn in the accident rate conversion model represents the accident rate, and y represents the adjustment factor of the relevant adjustment rules.

[0059] In the second aspect of the present invention, a risk assessment system for autonomous driving of intelligent connected vehicles is proposed. Figure 3 , including: a data collection module 10, configured to obtain the basic configuration and historical driving data of the target vehicle; a safety performance scoring module 20, configured to input the basic configuration into an autonomous driving safety performance scoring model pre-built based on multiple road scenarios to obtain a comprehensive safety performance score of the target vehicle; a disengagement rate prediction module 30, configured to input the safety performance score and historical driving data into a dynamic disengagement model pre-built based on driving behavior to obtain the disengagement rate of autonomous driving; an accident rate prediction module 40, configured to input the disengagement rate into an accident rate conversion model pre-built based on adjustment rules to obtain the accident rate of the target vehicle.

[0060] Among them, the data collection module 10, the safety performance scoring module 20, the escape rate prediction module 30 and the accident rate prediction module 40 are also configured to implement the steps in the above-mentioned corresponding method embodiments.

[0061] Through the above-mentioned intelligent connected vehicle automatic driving risk assessment system method or system, the present invention realizes a comprehensive assessment of the automatic driving performance of an intelligent connected vehicle, and can reasonably and accurately realize the risk assessment of the connected vehicle by predicting the disengagement rate of automatic driving and converting it into an accident rate.

[0062] The above are exemplary embodiments disclosed in the present invention, but it should be noted that various changes and modifications may be made without departing from the scope disclosed in the embodiments of the present invention as defined in the claims. The functions, steps and / or actions of the method claims according to the disclosed embodiments described herein do not need to be performed in any particular order. In addition, although the elements disclosed in the embodiments of the present invention may be described or required in individual form, they may also be understood as multiple unless explicitly limited to the singular.

[0063] The serial numbers of the embodiments disclosed in the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0064] A person skilled in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the disclosure of the embodiments of the present invention (including the claims) is limited to these examples; under the concept of the embodiments of the present invention, the technical features in the above embodiments or different embodiments can also be combined, and there are many other changes in different aspects of the above embodiments of the present invention, which are not provided in detail for the sake of simplicity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present invention should be included in the protection scope of the embodiments of the present invention.

Claims

1. A method for risk assessment of autonomous driving of intelligent connected vehicles, characterized in that: The method comprises: Obtain the basic configuration and historical driving data of the target vehicle; The basic configuration is input into an autonomous driving safety performance scoring model pre-built based on a variety of road scenarios to obtain a comprehensive safety performance score of the target vehicle; wherein the autonomous driving safety performance scoring model includes a primary weight corresponding to a primary scoring indicator, a secondary weight factor corresponding to a secondary scoring indicator, a secondary safety score, and a corresponding basic configuration combination, wherein the primary scoring indicator is obtained according to the safety performance classification of the vehicle's autonomous driving level, the secondary scoring indicator is obtained by dividing each of the primary scoring indicators according to a plurality of road scenarios, and the basic configuration combination is used as a matching sub-model to match the input static basic configuration to determine the corresponding secondary safety score and the corresponding secondary weight factor; Inputting the comprehensive safety performance score and the historical driving data into a dynamic disengagement model pre-constructed based on driving behavior to obtain an autonomous driving disengagement rate; wherein the dynamic disengagement model includes an autonomous driving key disengagement classification model and a corresponding classification influence weight, wherein the classification influence weight is determined by a preset score range of the comprehensive safety performance score, and the autonomous driving key disengagement classification model is used to predict the number of disengagement points based on key driving behaviors in the input autonomous driving historical driving data; The escape rate is input into an accident rate conversion model established in advance based on adjustment rules to obtain the accident rate of the target vehicle; wherein the accident rate conversion model includes a plurality of escape rate-accident rate conversion curves at different average speeds and corresponding adjustment factors, and the adjustment factors are determined by the number of accidents and the total compensation amount in the previous year, the city where the vehicle is located and / or the commonly used driving routes.

2. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to claim 1, characterized in that: The method further includes constructing the autonomous driving safety performance scoring model through the following steps, including: The safety performance of autonomous driving is divided into multiple first-level scoring indicators according to the vehicle's autonomous driving level; Dividing each of the first-level scoring indicators into a plurality of second-level scoring indicators according to a plurality of road scenes; Determine a basic configuration combination related to the secondary scoring indicator, wherein the basic configuration combination includes a basic hardware configuration and / or a basic software configuration; Testing the secondary safety score of the basic configuration combination under the secondary scoring index, and storing the secondary safety score in association with the basic configuration combination; Designing a secondary weight factor according to the basic hardware configuration level, quantity and / or whether there is a basic software configuration corresponding to the secondary scoring indicator in the basic configuration combination; Designing a first-level weight for each first-level scoring indicator according to the importance of the first-level scoring indicator to the safety performance of the vehicle's autonomous driving; The autonomous driving safety performance scoring model is constructed based on the first-level weight, the second-level weight factor, the second-level safety score and the corresponding basic configuration combination.

3. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to claim 2, characterized in that: The safety performance of autonomous driving is divided into multiple first-level scoring indicators according to the vehicle's autonomous driving level, including: The safety performance of autonomous driving is divided into vehicle following capability, combined control capability, emergency avoidance capability, and monitoring interaction capability.

4. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to claim 3, characterized in that: Dividing each of the first-level scoring indicators into a plurality of second-level scoring indicators according to a plurality of road scenes comprises: The following vehicle capability is divided into the following types: the preceding vehicle is stationary, the preceding vehicle is at a low speed, the preceding vehicle is decelerating, the preceding vehicle is cutting in, the preceding vehicle is cutting out, and / or starting and stopping following the preceding vehicle; The combined control capability is divided into lane centering control, traffic jam assistance, high-speed driving assistance, lane change without interference and / or lane change with interference; The emergency avoidance capability is divided into obstructed pedestrian crossing, nighttime pedestrian crossing, two-wheeled vehicle crossing, accident vehicle recognition and response, and / or cone barrel recognition and response; The monitoring interaction capability is divided into driver eyes-closed monitoring, driver head-down monitoring, driver status monitoring interaction and / or alarm event monitoring interaction.

5. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to any one of claims 2 to 4, characterized in that: The basic configuration is input into an autonomous driving safety performance scoring model pre-built based on a variety of road scenarios to obtain a comprehensive safety performance score of the target vehicle, including: Extracting keywords according to the functional description of the basic configuration as classification labels of the basic configuration; Clustering the classification labels according to the primary scoring index and the secondary scoring index and acquiring basic configurations corresponding to the classification labels under the same cluster to form a basic configuration combination; Determine the basic hardware configuration level, quantity and / or whether there is a basic software configuration corresponding to the secondary scoring indicator in the basic configuration combination and compare with the basic configuration combination information stored in the autonomous driving safety performance scoring model to determine the secondary weight; Determine a primary weight according to the primary scoring index corresponding to the configuration combination; Calculate the safety performance score of the target vehicle in each road scenario according to the primary weight, the secondary weight and the secondary safety score; The average safety performance score of the target vehicle in all road scenarios is calculated to obtain the comprehensive safety performance score.

6. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to claim 1, characterized in that: The method further comprises constructing the dynamic detachment model by the following steps, comprising: Acquire historical driving data of the target vehicle, analyze and extract key driving behaviors that affect the safety performance of autonomous driving from the historical driving data; During the test, record the first time point of critical disengagement; Marking the key driving behaviors extracted within a preset time before and after the first time point and using them as training samples; Training an autonomous driving key disengagement classification model based on the training samples to obtain an initial dynamic disengagement model; Different classification influence weights are configured for the initial dynamic disengagement model according to a preset score range to construct the dynamic disengagement model.

7. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to claim 6, characterized in that: Inputting the comprehensive safety performance score and the historical driving data into a dynamic disengagement model pre-built based on driving behavior to obtain a disengagement rate of autonomous driving, including: Determine the classification influence weight of the dynamic disengagement model according to the preset score range of the comprehensive safety performance score to determine the dynamic disengagement model used this time; Acquire historical driving data of the target vehicle and extract key driving behaviors; inputting the key driving behavior into the dynamic disengagement model to determine a disengagement point in the key driving behavior; The driving mileage at different average speeds is counted, and the disengagement rate of the autonomous driving is obtained by dividing the number of disengagement points occurring within the corresponding driving mileage by the corresponding driving mileage.

8. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to claim 7, characterized in that: The acquiring of the historical driving data of the target vehicle and the extraction of key driving behaviors include: The average speed of the target vehicle exceeding a specified speed threshold, the average acceleration exceeding a preset acceleration threshold, the sudden braking and / or the sharp turning are extracted.

9. The method for risk assessment of autonomous driving of an intelligent networked vehicle according to claim 1, characterized in that: The method further comprises constructing a risk rate conversion model through the following steps, including: Collect the data of the insured vehicles’ escape rate before the accident, and analyze the escape rate-accident rate conversion curve at different average speeds; Configuring an increaseable and modifiable adjustment factor for the escape rate-risk rate conversion curve, each of the adjustment factors corresponding to an adjustment rule; configuring a callback function for each of the adjustment factors to adjust the adjustment factor when the adjustment rule changes; Among them, the factors involved in the adjustment rules include: the number of accidents and the total compensation amount in the previous year, the city where the driver is located and / or the commonly used driving routes.

10. A risk assessment system for autonomous driving of intelligent connected vehicles, characterized in that: include: A data collection module configured to obtain basic configuration and historical driving data of a target vehicle; A safety performance scoring module, configured to input the basic configuration into an autonomous driving safety performance scoring model pre-built based on multiple road scenarios to obtain a comprehensive safety performance score of the target vehicle; wherein the autonomous driving safety performance scoring model includes a primary weight corresponding to a primary scoring indicator, a secondary weight factor corresponding to a secondary scoring indicator, a secondary safety score, and a corresponding basic configuration combination, wherein the primary scoring indicator is obtained according to the safety performance classification of the vehicle's autonomous driving level, the secondary scoring indicator is obtained by dividing each of the primary scoring indicators according to multiple road scenarios, and the basic configuration combination is used as a matching sub-model to match the input static basic configuration to determine the corresponding secondary safety score and the corresponding secondary weight factor; a disengagement rate prediction module, configured to input the safety performance score and the historical driving data into a dynamic disengagement model pre-built based on driving behavior to obtain an autonomous driving disengagement rate; wherein the dynamic disengagement model includes an autonomous driving key disengagement classification model and a corresponding classification influence weight, the classification influence weight is determined by a preset score range of the comprehensive safety performance score, and the autonomous driving key disengagement classification model is used to predict the number of disengagement points based on key driving behaviors in the input autonomous driving historical driving data; An accident rate prediction module is configured to input the escape rate into an accident rate conversion model established in advance based on adjustment rules to obtain the accident rate of the target vehicle; wherein the accident rate conversion model includes escape rate-accident rate conversion curves under multiple average speeds and corresponding adjustment factors, and the adjustment factors are determined by the number of accidents and total compensation amount in the previous year, the city where the vehicle is located and / or the commonly used driving route.

Images