An accident cause determination method, device, equipment and storage medium

Abstract

An embodiment of the present invention discloses a method, apparatus, device, and storage medium for determining the cause of an accident. The method includes: obtaining a set of accident scenarios matching a target type of accident; wherein, the accident scenarios are marked with scenario feature tags; establishing a scenario decision tree from the scenario features to the accident severity according to the set of accident scenarios; and determining the cause of the target type of accident scenario to be processed according to the scenario decision tree. The technical solution of the embodiment of the present invention solves the problem in the related art that the cause of an autonomous driving accident cannot be effectively analyzed, and realizes analyzing the cause of a tagged autonomous driving accident scenario according to the accident scenario decision tree.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method, device, equipment and storage medium for determining the cause of an accident. Background Art

[0002] At present, although many countries are vigorously developing autonomous driving technology and promoting large-scale testing of complex and diverse traffic scenarios, the safety risks of autonomous driving have not yet been effectively prevented and controlled. Traffic accidents involving various types of autonomous driving vehicles are common. How to analyze accident scenarios and determine the causes of accidents has become one of the research hotspots in the field of autonomous driving. Summary of the Invention

[0003] Embodiments of the present invention provide a method, apparatus, device, and storage medium for determining the cause of an accident to solve the problem in related technologies that the cause of an autonomous driving accident cannot be effectively analyzed, and realize the analysis of the cause of the accident in a labeled autonomous driving accident scenario based on an accident scenario decision tree.

[0004] According to one aspect of the present invention, a method for determining the cause of an accident is provided, comprising:

[0005] Obtain a set of accident scenes that match the target accident; wherein the accident scenes are marked with scene feature labels;

[0006] Based on the accident scenario set, a scenario decision tree is established which points to the severity of the accident from the scenario features;

[0007] According to the scenario decision tree, determine the cause of the target accident scenario to be processed.

[0008] Optionally, obtain a set of accident scenarios that match the target type of accident, including:

[0009] Get the target accident type corresponding to the target accident;

[0010] From the preset labeled accident scenes, select the accident scenes marked with the target accident type label and add them to the accident scene collection.

[0011] Optionally, based on the accident scenario set, a scenario decision tree is established that points from scenario features to accident severity, including:

[0012] According to the target accident severity corresponding to the target type accident, the information gain value of each scenario feature under the accident scenario set is calculated to determine the root decision node of the scenario decision tree;

[0013] The root decision node is used as the current processing node, and according to the remaining scene features, the decision node or leaf node based on each feature attribute of the current processing node is determined;

[0014] Take each decision node as the current processing node, and return the operation of executing and determining the decision node or leaf node of each feature attribute based on the current processing node according to the remaining scenario features until all leaf nodes point to the accident severity.

[0015] Optionally, calculate the information gain value of each scenario feature under the accident scenario set according to the target accident severity corresponding to the target class of accidents, and determine the root decision node of the scenario decision tree, including:

[0016] Based on the accident scenario set, calculate the information entropy of the target class of accidents according to the target accident severity corresponding to the target class of accidents;

[0017] Calculate the conditional entropy of the target class of accidents based on each scenario feature;

[0018] Calculate the difference between the information entropy and the conditional entropy corresponding to each scenario feature as the information gain value of each scenario feature;

[0019] Select the scenario feature with the largest information gain value as the root decision node of the scenario decision tree.

[0020] Optionally, determine the decision node or leaf node of each feature attribute based on the current processing node according to the remaining scenario features, including:

[0021] For each feature attribute of the current processing node, filter out the accident scenario subset corresponding to each feature attribute from the accident scenario set;

[0022] For each accident scenario subset, calculate the information entropy of the target class of accidents based on the corresponding feature attribute according to the target accident severity corresponding to the target class of accidents;

[0023] If the information entropy is zero, determine the leaf node pointed to by the feature attribute; the leaf node is the accident severity;

[0024] If the information entropy is not zero, determine the decision node based on the feature attribute from the remaining scenario features.

[0025] Optionally, determine the accident cause of the target class of accident scenarios to be processed according to the scenario decision tree, including:

[0026] Perform labeling processing on the target class of accident scenarios to determine the first scenario feature label corresponding to the target class of accident scenarios;

[0027] According to the scenario decision tree, query the first node path matching the first scenario feature label;

[0028] Sort each first scenario feature according to the first node path to obtain the accident cause of the target class of accident scenarios.

[0029] Optionally, the scene features at least include: weather, lighting, vehicle speed limit, number of lanes, slope, curvature, and intersection type.

[0030] According to another aspect of the present invention, there is provided an accident cause determination device, including:

[0031] An accident scene acquisition module, configured to execute acquiring a set of accident scenes matching a target type of accident; wherein, the accident scenes are marked with scene feature tags;

[0032] A scene decision tree establishment module, configured to execute establishing a scene decision tree pointing from scene features to accident severity according to the set of accident scenes;

[0033] An accident cause analysis module, configured to execute determining the accident cause of a target type of accident scene to be processed according to the scene decision tree. According to another aspect of the present invention, there is provided an electronic device, which includes:

[0034] At least one processor; and

[0035] A memory communicatively connected to the at least one processor; wherein,

[0036] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the accident cause determination method of any embodiment of the present invention.

[0037] According to another aspect of the present invention, there is provided a computer-readable storage medium, which stores computer instructions for causing a processor to implement the accident cause determination method of any embodiment of the present invention when executed.

[0038] The technical solution of the embodiment of the present invention solves the problem in the related art that the cause of an autonomous driving accident cannot be effectively analyzed by acquiring a set of accident scenes matching a target type of accident; wherein, the accident scenes are marked with scene feature tags; establishing a scene decision tree pointing from scene features to accident severity according to the set of accident scenes; and determining the accident cause of a target type of accident scene to be processed according to the scene decision tree, and achieves the beneficial effect of analyzing the accident cause of a labeled autonomous driving accident scene according to the accident scene decision tree.

[0039] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0041] Figure 1 is a flowchart of a method for determining the cause of an accident provided in Embodiment 1 of the present invention;

[0042] Figure 2 is a flowchart of a method for determining the cause of an accident provided in Embodiment 2 of the present invention;

[0043] Figure 3 is a schematic diagram of a scenario decision tree applicable to Embodiment 2 of the present invention;

[0044] Figure 4 is a schematic structural diagram of an apparatus for determining the cause of an accident provided in Embodiment 3 of the present invention;

[0045] Figure 5 is a schematic structural diagram of an electronic device for implementing the method for determining the cause of an accident in the embodiments of the present invention. Detailed Embodiments

[0046] In order to enable those skilled in the art to better understand the solutions of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] It should be noted that the terms "target", "first", etc. in the specification and claims of the present invention and the above accompanying drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily need to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0048] Embodiment 1

[0049] Figure 1It is a flowchart of a method for determining the cause of an accident provided in the first embodiment of the present invention. This embodiment is applicable to the situation of analyzing the cause of an autonomous driving accident scenario. This method can be executed by an accident cause determination device, which can be implemented in the form of hardware and / or software, and this device can be configured in an electronic device. As Figure 1 shown, the method includes:

[0050] S110. Obtain a set of accident scenarios that match the target type of accident.

[0051] Among them, an accident scenario is a file describing the situation of an autonomous driving accident, and it can be obtained through autonomous driving simulation according to pre-set road files, environmental files, the driving situation file of the vehicle itself, the driving situation files of other vehicles, etc. Exemplarily, the road file may include parameters such as road direction, road surface environment, number of lanes, slope, presence of obstacles, etc.; the environmental file may include weather, light, etc.; the driving situation file of the vehicle itself and the driving situation files of other vehicles may include parameters such as vehicle driving route, driving mode, driving speed, vehicle health status, etc.

[0052] Among them, the accident scenario is marked with a scenario feature label. A scenario feature is a possible cause of an accident, and the value of the scenario feature label can be a feature attribute under this scenario feature. In an implementable manner, the scenario features at least include: weather, light, speed limit, number of lanes, slope, curvature, and intersection type. Exemplarily, the weather label marked on accident scenario A is sunny, the light label is daylight, the speed limit label is <40 km / h, the number of lanes label is 3 lanes, the slope label is low slope (≤4%), the curvature label is small curvature (curvature radius >500 m), and the intersection type label is roundabout.

[0053] Among them, the target type of accident can be any type of autonomous driving accident, such as a major frontal collision accident, a minor rear-end collision accident, etc. In this embodiment, by combining the accident type and the severity of the accident, the accident category of a certain autonomous driving accident scenario is determined. For example, if the accident type of an accident scenario is frontal collision and the severity of the accident is a major accident, then the accident category of this accident scenario is a major frontal collision accident.

[0054] Optionally, obtaining a set of accident scenarios that match the target type of accident includes: obtaining the target accident type corresponding to the target type of accident; selecting the accident scenarios marked with the target accident type label from the pre-set labeled accident scenarios and adding them to the set of accident scenarios.

[0055] In this embodiment, a large number of accident scenarios are pre-tagged, and accident type tags, accident severity tags, and scenario feature tags are added to each accident scenario. When performing accident causation analysis on a target type of accident, accident scenarios with the same accident type as the target type of accident can be screened out from a large number of tagged accident scenarios to form an accident scenario set. For example, when analyzing the cause of a particularly serious accident caused by a frontal collision, accident scenarios with the accident type tag of frontal collision need to be screened out to form a corresponding accident scenario set for establishing a scenario decision tree.

[0056] In this embodiment, the accident scenarios can be tagged according to the classification rules shown in Table 1. Among them, the types of autonomous driving accidents can be divided into four categories: frontal collision, rear-end collision, side collision, and single-motor vehicle accident. The accident causes can be jointly determined by 7 accident scenario feature tags such as weather, lighting, speed limit, number of lanes, slope, curvature, and intersection type in Table 1, and finally point to the accident severity in Table 2, including minor accidents, general accidents, major accidents, and particularly serious accidents.

[0057] Table 1 Method for Tagging Accident Scenarios

[0058]

[0059] Table 2 Classification Criteria for Accident Severity

[0060]

[0061] In this embodiment, by tagging the accident scenarios, the classification and feature extraction of the accident scenarios are completed, and the originally complex accident scenarios are represented only by 9 tags, greatly reducing the analysis difficulty.

[0062] S120. Establish a scenario decision tree that points from scenario features to accident severity according to the accident scenario set.

[0063] In this embodiment, in order to analyze the influence degree of each scenario feature on the cause of the target type of accident, by analyzing the accident scenarios matched with the target type of accident, a scenario decision tree that points from scenario features to accident severity is constructed. The role of the scenario decision tree is equivalent to a classifier, which can be used to analyze the accident causation of a certain type or a certain accident scenario, and sort all the scenario features that may cause an accident according to the degree of influence on the accident. As Figure 2 shown is a structure of the scenario decision tree. Of course, the structure of the scenario decision tree mainly depends on the number of feature attributes corresponding to each decision node and is not limited to the binary tree structure.

[0064] Among them, when establishing the scenario decision tree for a certain type of accident scenario, the scenario features corresponding to the root decision node are the main causes of the occurrence of this type of accident. The scenario features corresponding to each decision node are sorted in order of importance and finally point to the leaf node, that is, the accident severity. For example, when analyzing the accident causes of a particularly serious frontal collision accident, according to the pre-set calculation formula, the root decision node and each decision node are determined, that is, the degree of accident causation of scenario features such as weather and lighting is calculated. Finally, the scenario decision tree will point to the accident severity, that is, particularly serious accidents and non-particularly serious accidents.

[0065] Optionally, according to the accident scenario set, establish a scenario decision tree that points from scenario features to accident severity, including: calculating the information gain value of each scenario feature under the accident scenario set according to the target accident severity corresponding to the target type of accident, and determining the root decision node of the scenario decision tree; taking the root decision node as the current processing node, and determining the decision node or leaf node based on each feature attribute of the current processing node according to the remaining scenario features; taking each decision node as the current processing node, and returning to execute the operation of determining the decision node or leaf node based on each feature attribute of the current processing node according to the remaining scenario features until all leaf nodes point to the accident severity.

[0066] In this embodiment, according to the target accident severity corresponding to the target type of accident, the accident scenarios in the accident scenario set are classified. For example, they are divided into particularly serious accident scenarios and non-particularly serious accident scenarios, and the information entropy of the target type of accident is calculated. Then calculate the conditional entropy corresponding to each scenario feature to obtain the information gain value. Select the scenario feature with the largest information gain value as the root decision node. After determining the root decision node, branch according to the feature attributes of the root decision node, and under each branch, repeat the above process to find the scenario feature with the largest information gain value as the decision node from the other scenario features except the root decision node, or find the accident severity as the leaf node. For example, if the root decision node is determined to be the weather, then branch outwards according to the feature attributes of the weather (sunny, rainy, foggy, windy, cloudy, snowy, other weather) and connect to the decision node. The decision node can be any other scenario feature except the weather reason. Determine the root decision node and each decision node according to the above process and finally point to the leaf node, such as particularly serious accidents and non-particularly serious accidents, to complete the establishment of the scenario decision tree for the target type of accident.

[0067] Among them, if analyzing other types of accident scenarios, such as minor frontal collision accidents, general frontal collision accidents, major frontal collision accidents, etc., only need to repeat the above steps to establish the corresponding accident scenario decision tree.

[0068] S130. Determine the accident cause of the target type of accident scenario to be processed according to the scenario decision tree.

[0069] In this embodiment, after establishing a scenario decision tree corresponding to a certain type of accident, the scenario decision tree can be used to analyze the accident causes of accident scenarios of the same type, and all scenario features that may cause the accident can be sorted according to the degree of impact on the accident as the accident causes. At the same time, it can also guide developers and testers to carry out targeted optimization for the main causes, reduce the safety risks of the autonomous driving system, reduce the number of casualties in traffic accidents, and improve the safety and reliability of the autonomous driving system.

[0070] Optionally, determining the accident cause of the target class of accident scenarios to be processed according to the scenario decision tree includes: performing tagging processing on the target class of accident scenarios to determine the first scenario feature tags corresponding to the target class of accident scenarios; querying the first node path that matches the first scenario feature tags according to the scenario decision tree; sorting each first scenario feature according to the first node path to obtain the accident cause of the target class of accident scenarios.

[0071] In this embodiment, after obtaining the target class of accident scenarios whose causes are to be analyzed, if there are no marked scenario feature tags in the accident scenario, the accident scenario is tagged according to the classification rules set in Table 1 and Table 2 to determine the first scenario feature tags corresponding to the accident scenario; if the accident scenario has been tagged, the first scenario feature tags corresponding to the accident scenario are directly obtained. For example, the first scenario feature tags include: rainy day, low slope (≤4%), dim, 3 lanes, 50 - 60 km / h, small curvature (radius of curvature > 500 m), roundabout. Use the first scenario feature tags to match each node path in the scenario decision tree, find the matching first node path, and sort each first scenario feature according to its position in the first node path as the accident cause of the accident scenario. For example, rainy day, 50 - 60 km / h, roundabout, dim, 3 lanes, small curvature (radius of curvature > 500 m), low slope (≤4%).

[0072] The technical solution of the embodiment of the present invention solves the problem in the related art that the causes of autonomous driving accidents cannot be effectively analyzed by obtaining a set of accident scenarios matching the target class of accidents; wherein, the accident scenarios are marked with scenario feature tags; establishing a scenario decision tree pointing from the scenario features to the accident severity according to the set of accident scenarios; and determining the accident cause of the target class of accident scenarios to be processed according to the scenario decision tree, and realizes analyzing the accident causes of the tagged autonomous driving accident scenarios according to the accident scenario decision tree.

[0073] Embodiment Two

[0074] Figure 2It is a flowchart of a method for determining the cause of an accident provided in Embodiment 2 of the present invention. On the basis of the above embodiments, this embodiment further provides: specific steps for calculating the information gain value of each scenario feature under the accident scenario set according to the target accident severity corresponding to the target type of accident, determining the root decision node of the scenario decision tree, and, according to the remaining scenario features, determining the decision nodes or leaf nodes of each feature attribute based on the current processing node. As Figure 2 shown, the method includes:

[0075] S210. Obtain an accident scenario set that matches the target type of accident.

[0076] Among them, the accident scenarios are marked with scenario feature labels.

[0077] S220. Calculate the information gain value of each scenario feature under the accident scenario set according to the target accident severity corresponding to the target type of accident, and determine the root decision node of the scenario decision tree.

[0078] In this embodiment, since all the accident scenarios in the accident scenario set correspond to the target accident type, for example, the frontal collision type, and the accident severity is divided into two types: target accident severity and non-target accident severity, for example, extremely serious accident and non-extremely serious accident. Therefore, the accident severity can be used as a random variable to calculate the information entropy of the target type of accident, that is, the uncertainty. Furthermore, according to the accident scenario set, calculate the conditional entropy of the target type of accident under the condition that each scenario feature occurs, determine the information gain value corresponding to each scenario feature, and use the scenario feature with the largest information gain value as the main cause of the accident, that is, the root decision node.

[0079] Optionally, calculating the information gain value of each scenario feature under the accident scenario set according to the target accident severity corresponding to the target type of accident and determining the root decision node of the scenario decision tree includes: calculating the information entropy of the target type of accident based on the accident scenario set according to the target accident severity corresponding to the target type of accident; calculating the conditional entropy of the target type of accident based on each scenario feature; calculating the difference between the information entropy and the conditional entropy corresponding to each scenario feature as the information gain value of each scenario feature; selecting the scenario feature with the largest information gain value as the root decision node of the scenario decision tree.

[0080] In this embodiment, when selecting the root decision node, the accident scenarios can be first divided into two categories according to whether they correspond to the target accident severity, so as to calculate the probabilities of the random variable Y taking the values of the target accident severity and the non-target accident severity respectively. Taking the example of a head-on collision with an extremely serious accident, the accident severity Y is divided into two cases. One is an extremely serious accident (denoted as y1), and the other is a non-extremely serious accident (denoted as y2). Then the probability p(y1) that Y takes the value of an extremely serious accident is equal to the ratio of the number of extremely serious accident scenarios to the total number of accident scenarios, and the probability p(y2) that Y takes the value of a non-extremely serious accident is equal to the ratio of the number of non-extremely serious accident scenarios to the total number of accident scenarios. Substitute p(y1) and p(y2) into the formula The information entropy H(Y) of the target type of accident can be calculated

[0081] Then, according to the formula Calculate the information entropy of the random variable Y on the premise that the accident cause is the scenario feature X, that is, the conditional entropy of the target type of accident based on the scenario feature X. Among them, x is any one of the feature attributes in the scenario feature X, p(x) is the probability of the occurrence of the feature attribute x, p(y1|x) is the probability of the occurrence of the random variable y1 under the condition that the feature attribute x occurs, and p(y2|x) is the probability of the occurrence of the random variable y2 under the condition that the feature attribute x occurs. Taking the head-on collision with an extremely serious accident as an example, the scenario feature X can be any one of weather, lighting, speed limit, number of lanes, slope, curvature, etc. When the scenario feature X is weather, calculate the information entropy of sunny day x1, rainy day x2, foggy day x3, windy day x4, cloudy day x5, snowy day x6, and other weather x7 respectively and accumulate them as the conditional entropy of the accident cause being weather and a head-on collision with an extremely serious accident occurring

[0082] Then, according to the formula Gain(Y,X) = H(Y) - H(Y|X), calculate the information entropy H(Y|X) of the random variable Y on the premise that the accident cause is determined to be the scenario feature X, and the information entropy H(Y) when the accident cause is not determined by any scenario feature, and take the difference between the two as the information gain value of the scenario feature X. Taking the head-on collision with an extremely serious accident as an example, this step calculates how much the entropy value of a head-on collision with an extremely serious accident decreases when the scenario feature X is determined to be weather

[0083] Repeat the above process to calculate the information gain values of the scenario features X being weather, lighting, speed limit, number of lanes, slope, curvature, and intersection type respectively. Since the influence degree of the scenario feature on the accident cause is the greatest when the information gain value is the largest, select the scenario feature with the largest information gain value as the root decision node

[0084] S230. Take the root decision node as the current processing node, and determine the decision nodes or leaf nodes based on the respective feature attributes of the current processing node according to the remaining scenario features.

[0085] In this embodiment, after determining the root decision node, branch according to the feature attributes of the root decision node, and under each branch, repeat the above process to find the scenario feature with the largest information gain value from the other scenario features except the root decision node as the decision node, or find the accident severity as the leaf node.

[0086] Optionally, determining the decision nodes or leaf nodes based on the respective feature attributes of the current processing node according to the remaining scenario features includes: for each feature attribute of the current processing node, screen out the sub-sets of accident scenarios corresponding to the respective feature attributes from the accident scenario set; for each sub-set of accident scenarios, calculate the information entropy of the target type of accident based on the corresponding feature attribute according to the target accident severity corresponding to the target type of accident; if the information entropy is zero, determine the leaf node pointed to by the feature attribute; the leaf node is the accident severity; if the information entropy is not zero, determine the decision node based on the feature attribute from the remaining scenario features.

[0087] In this embodiment, for each feature attribute of the current processing node, select the accident scenarios marked with the corresponding labels from the accident scenario set respectively to form the corresponding sub-sets of accident scenarios. Then, according to the sub-sets of accident scenarios corresponding to each feature attribute, calculate the information gain value of each scenario feature except the current processing node on the premise that the main accident cause is determined to be this feature attribute, and select the scenario feature with the largest information gain value as the decision node based on this feature attribute, that is, the secondary accident cause when the main accident cause is this feature attribute.

[0088] Exemplarily, for a particularly serious frontal collision accident, assume that it has been calculated that the root decision node is the weather, and the root decision node is determined as the current processing node. Then, for each feature attribute of the weather, such as rainy day, select the accident scenarios marked with the rainy day label from the accident scenario set corresponding to the current processing node to form the sub-set of accident scenarios corresponding to the rainy day. Then, classify the sub-set of accident scenarios according to whether it is a particularly serious accident, calculate the probability of a particularly serious accident occurring on the premise of a rainy day and the probability of a non-particularly serious accident occurring on the premise of a rainy day, and substitute them into the information entropy calculation formula to calculate the information entropy of a particularly serious frontal collision accident occurring when the main accident cause is a rainy day.

[0089] If the information entropy is zero, that is, all accident scenarios labeled with the rainy day tag are of the same accident severity level, such as extremely serious accidents or non-extremely serious accidents, then determine that the rainy day attribute of the current processing node points to the leaf node of the scenario decision tree, and use the accident severity level at this time as the leaf node. If the information entropy is not zero, then calculate the conditional entropy of a frontal collision extremely serious accident when the secondary accident cause is other scenario feature X except weather on the premise that the main accident cause is rainy day, and then calculate the information gain value of the remaining scenario features on the premise of determining rainy day. Select the scenario feature with the largest information gain value as the decision node under the rainy day branch of the root decision node. Repeat the above process to calculate the decision nodes or leaf nodes of the weather under the branches of each feature attribute.

[0090] S240. Take each decision node as the current processing node, and return to execute the operation of determining the decision node or leaf node of each feature attribute based on the current processing node according to the remaining scenario features until all leaf nodes point to the accident severity level.

[0091] S250. Determine the accident cause of the target class accident scenario to be processed according to the scenario decision tree.

[0092] In this embodiment, in the accident scenario decision tree, each scenario feature is sorted according to the degree of influence on the accident, and finally points to the accident severity level. According to the accident scenario decision trees corresponding to various types of accidents, it can not only guide developers and testers to optimize the accident causes for a certain type or even a certain scenario specifically, but also predict whether an accident will occur in certain scenarios.

[0093] Optionally, it further includes: in response to an accident prediction operation for a specified scenario, performing a tagging process on the specified scenario to determine a second scenario feature tag corresponding to the specified scenario; performing node path matching on the scenario decision trees corresponding to various types of accidents according to the second scenario feature tag; if there is a second node path that matches the second scenario feature tag, then determine that an accident will occur in the specified scenario.

[0094] In this embodiment, when predicting whether an accident will occur in the future for a certain scenario, scenario feature tags such as weather, light, speed limit, number of lanes, slope, and curvature can be marked on the scenario first, that is, the second scenario feature tag. Match the scenario decision trees corresponding to various types of accidents according to the second scenario feature tag to determine whether a second node path that matches the second scenario feature tag can be found. If not, it is initially considered that no accident will occur in this scenario; if it can be found, it is considered that an accident will occur in this scenario in the future, and the accident severity level is the leaf node of the second node path, and the accident type is the accident type corresponding to this scenario decision tree.

[0095] In this embodiment, by setting the calculation rules of the root decision node and the decision nodes, the accident impact degrees are sorted according to various scenario features, and an accident scenario decision tree is established. According to the accident scenario decision tree, not only can the accident causes of a certain type or even a certain accident scenario be pointed out, but also the accident prediction can be carried out for some scenarios where accidents have not occurred.

[0096] Embodiment III

[0097] Figure 4 As shown in the structural schematic diagram of an accident cause determination device provided in Embodiment III of the present invention. Figure 4 As shown in the figure, the device includes: an accident scenario acquisition module 310, a scenario decision tree establishment module 320, and an accident cause analysis module 330.

[0098] The accident scenario acquisition module 310 is configured to execute acquiring an accident scenario set matching a target type of accident; wherein, the accident scenarios are marked with scenario feature labels.

[0099] The scenario decision tree establishment module 320 is configured to execute establishing a scenario decision tree pointing from the scenario features to the accident severity according to the accident scenario set.

[0100] The accident cause analysis module 330 is configured to execute determining the accident cause of the target type of accident scenario to be processed according to the scenario decision tree.

[0101] Optionally, the accident scenario acquisition module 310 is configured to execute:

[0102] acquiring a target accident type corresponding to the target type of accident;

[0103] selecting the accident scenarios marked with the target accident type label from the preset labeled accident scenarios and adding them to the accident scenario set.

[0104] Optionally, the scenario decision tree establishment module 320 includes:

[0105] a root decision node determination unit configured to execute calculating the information gain values of various scenario features under the accident scenario set according to the target accident severity corresponding to the target type of accident, and determining the root decision node of the scenario decision tree;

[0106] a decision node determination unit configured to execute using the root decision node as the current processing node, and determining the decision nodes or leaf nodes of each feature attribute based on the current processing node according to the remaining scenario features;

[0107] a return unit configured to execute using each decision node as the current processing node, and returning to execute the operation of determining the decision nodes or leaf nodes of each feature attribute based on the current processing node according to the remaining scenario features until all leaf nodes point to the accident severity.

[0108] Optionally, a root decision node determination unit is configured to execute:

[0109] Based on the accident scenario set, according to the target accident severity corresponding to the target type of accident, calculate the information entropy of the target type of accident;

[0110] Calculate the conditional entropy of the target type of accident based on various scenario features;

[0111] Calculate the difference between the information entropy and the conditional entropy corresponding to each scenario feature, and use it as the information gain value of each scenario feature;

[0112] Select the scenario feature with the largest information gain value as the root decision node of the scenario decision tree.

[0113] Optionally, a decision node determination unit is configured to execute:

[0114] For each feature attribute of the current processing node, filter out the accident scenario sub-sets corresponding to each feature attribute from the accident scenario set;

[0115] For each accident scenario sub-set, according to the target accident severity corresponding to the target type of accident, calculate the information entropy of the target type of accident based on the corresponding feature attribute;

[0116] If the information entropy is zero, determine the leaf node pointed to by the feature attribute; the leaf node is the accident severity;

[0117] If the information entropy is not zero, determine the decision node based on the feature attribute from the remaining scenario features.

[0118] Optionally, an accident cause analysis module 330 is configured to execute:

[0119] Perform labeling processing on the target type of accident scenario to determine the first scenario feature label corresponding to the target type of accident scenario;

[0120] According to the scenario decision tree, query the first node path that matches the first scenario feature label;

[0121] Sort each first scenario feature according to the first node path to obtain the accident cause of the target type of accident scenario.

[0122] Optionally, the scenario features at least include: weather, lighting, vehicle speed limit, number of lanes, slope, curvature, and intersection type.

[0123] The accident cause determination device provided by the embodiments of the present invention can execute the accident cause determination method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0124] Embodiment 4

[0125] Figure 5 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0126] As Figure 5 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0127] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0128] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the accident cause determination method.

[0129] In some embodiments, the accident cause determination method may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the accident cause determination method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the accident cause determination method by any other suitable means (e.g., by means of firmware).

[0130] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems on a chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0131] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer programs are executed by the processor, the functions / operations specified in the flowchart and / or block diagram are implemented. The computer programs can be executed entirely on the machine, partially on the machine, as a stand-alone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0132] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media would include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0133] To provide for interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide for interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0134] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0135] A computing system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The relationship between the client and the server is created by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0136] It should be understood that various forms of the processes shown above can be used, steps can be reordered, added or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.

[0137] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for determining the cause of an accident, characterized in that Including: Obtain a set of accident scenarios matching the target type of accident; among them, the accident scenarios are marked with scenario feature tags; Based on the set of accident scenarios, establish a scenario decision tree that points from scenario features to accident severity; Based on the scenario decision tree, determine the cause of the target type of accident scenario to be processed; The establishing a scenario decision tree that points from scenario features to accident severity based on the set of accident scenarios includes: Based on the target accident severity corresponding to the target type of accident, calculate the information gain value of each scenario feature under the set of accident scenarios, and determine the root decision node of the scenario decision tree; Take the root decision node as the current processing node, and based on the remaining scenario features, determine the decision nodes or leaf nodes of each feature attribute based on the current processing node; Take each of the decision nodes as the current processing node, and return to execute the operation of determining the decision nodes or leaf nodes of each feature attribute based on the current processing node according to the remaining scenario features until all leaf nodes point to accident severity.

2. The method according to claim 1, wherein The obtaining a set of accident scenarios matching the target type of accident includes: Obtain the target accident type corresponding to the target type of accident; From the preset labeled accident scenarios, select the accident scenarios marked with the target accident type label and add them to the set of accident scenarios.

3. The method according to claim 1, characterized in that, The calculating the information gain value of each scenario feature under the set of accident scenarios based on the target accident severity corresponding to the target type of accident and determining the root decision node of the scenario decision tree includes: Based on the set of accident scenarios, calculate the information entropy of the target type of accident according to the target accident severity corresponding to the target type of accident; Calculate the conditional entropy of the target type of accident based on each scenario feature; Calculate the difference between the information entropy and the conditional entropies corresponding to each scenario feature as the information gain value of each scenario feature; Select the scenario feature with the largest information gain value as the root decision node of the scenario decision tree.

4. The method according to claim 1, characterized in that, The determining the decision nodes or leaf nodes of each feature attribute based on the current processing node according to the remaining scenario features includes: For each feature attribute of the current processing node, filter out the sub-sets of accident scenarios corresponding to each feature attribute from the set of accident scenarios; For each sub-set of accident scenarios, calculate the information entropy of the target type of accident based on the corresponding feature attribute according to the target accident severity corresponding to the target type of accident; If the information entropy is zero, determine the leaf node pointed to by the feature attribute; the leaf node is the accident severity; If the information entropy is not zero, determine the decision node based on the feature attribute from the remaining scenario features.

5. The method according to claim 1, wherein The determining the cause of the target type of accident scenario to be processed based on the scenario decision tree includes: Perform a labeling process on the target type of accident scenario to determine the first scenario feature label corresponding to the target type of accident scenario; According to the scenario decision tree, query the first node path that matches the first scenario feature label; Sort each first scenario feature according to the first node path to obtain the accident cause of the target type of accident scenario.

6. The method according to any one of claims 1-5, characterized in that, The scenario features at least include: weather, lighting, vehicle speed limit, number of lanes, slope, curvature, and intersection type.

7. An accident cause determination device, characterized in that, It includes: An accident scenario acquisition module, configured to execute acquiring an accident scenario set matching a target type of accident; wherein, scenario feature labels are marked in the accident scenarios. A scenario decision tree establishment module, configured to execute establishing a scenario decision tree from scenario features to accident severity according to the accident scenario set. An accident cause analysis module, configured to execute determining the accident cause of the target type of accident scenario to be processed according to the scenario decision tree. The scenario decision tree establishment module includes: A root decision node determination unit, configured to execute calculating the information gain value of each scenario feature under the accident scenario set according to the target accident severity corresponding to the target type of accident, and determining the root decision node of the scenario decision tree. A decision node determination unit, configured to execute using the root decision node as the current processing node, and determining decision nodes or leaf nodes based on each feature attribute of the current processing node according to the remaining scenario features. A return unit, configured to execute using each of the decision nodes as the current processing node, and returning to execute the operation of determining decision nodes or leaf nodes based on each feature attribute of the current processing node according to the remaining scenario features, until all leaf nodes point to accident severity.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor, so that the at least one processor can execute the accident cause determination method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the processor to implement the accident cause determination method according to any one of claims 1-6 when executed.

Images