Analysis method and system for aided driving road test

Abstract

The invention provides an analysis method and system for an auxiliary driving road test. The method comprises the following steps: acquiring event information of a problem event occurring in an auxiliary driving test process; determining a problem phenomenon corresponding to the event information according to the event information; calling an analysis model corresponding to the problem phenomenon to analyze the event information to obtain an analysis result of the problem event; namely, the problem events recorded in the road test are analyzed, the situation that each problem event is analyzed manually through an analysis model is avoided, the time cost of manual data analysis is saved, meanwhile, before the analysis model is called, the problem phenomenon of the problem event is determined, and then the corresponding analysis model is called for targeted analysis, so that the analysis efficiency is improved. And the analysis accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent automobiles, and more specifically, relates to an analysis method and system for assisted driving road testing. Background Art

[0002] The test engineer is responsible for recording the road test problems of the assisted driving function, and gives the recorded time points and data related to the test problems to analysts for manual problem data analysis.

[0003] The problem recording process mainly relies on recording the actual vehicle performance of the assisted driving function in actual road conditions; this means that as long as the vehicle driven by the driver encounters any problems that cause driver discomfort, poor vehicle performance and subjective expectations during driving, they will be recorded, regardless of whether the problem is caused by functional abnormality or design that does not meet expectations; therefore, among the problems recorded during road testing, there are a large number of invalid problems that are not related to the function or are caused by functional abnormalities of non-assisted driving algorithms; therefore, subsequent analysis of these problems is particularly important.

[0004] In the existing technology, all these problems will be transferred to analysts for data analysis on the same day. Since invalid problems account for a large proportion, this results in a long analysis time, high labor costs, and limited value. Summary of the invention

[0005] In view of this, an object of the present invention is to provide an analysis method and system for assisted driving road testing, which are used to perform analysis using an analysis model to save the time cost of manual data analysis.

[0006] The first aspect of the present application discloses an analysis method for assisted driving road testing, comprising:

[0007] Obtain event information of problem events that occurred during assisted driving testing;

[0008] Determine, according to the event information, the problem phenomenon corresponding to the event information;

[0009] The analysis model corresponding to the problem phenomenon is called to analyze the event information to obtain the analysis result of the problem event.

[0010] Optionally, the calling of the analysis model corresponding to the problem to analyze the event information to obtain the analysis result of the problem event includes:

[0011] Taking the analysis model corresponding to the problem phenomenon as the target analysis model;

[0012] Calling the target analysis model, and inputting the vehicle model parameters corresponding to the event information into the target analysis model;

[0013] The target analysis model performs analysis processing to obtain an analysis result of the problem event.

[0014] Optionally, the analysis model performs analysis processing to obtain an analysis result of the problem event, including:

[0015] Acquire driving test data corresponding to the problem phenomenon within the positioning time corresponding to the problem phenomenon;

[0016] Based on the driving test data, logical judgment is performed on each parameter in the driving test data in sequence according to the structure of the link tree to obtain an analysis result.

[0017] Optionally, the step of performing logical judgment on each parameter in the driving test data in sequence according to the structure of a link tree based on the driving test data to obtain an analysis result includes:

[0018] Based on the driving test data, logical judgment is performed on each parameter in the driving test data in sequence according to the structure of the link tree to determine the analysis conclusion;

[0019] Matching corresponding solutions and problem types according to the analysis conclusions;

[0020] The analysis conclusion, solution and problem type are collectively referred to as the analysis result.

[0021] Optionally, determining, according to the event information, a problem phenomenon corresponding to the event information includes:

[0022] The event information is traversed using various problem phenomenon determination algorithms;

[0023] A method for determining whether the event information satisfies each of the problem phenomenon determination algorithms;

[0024] If the event information satisfies any of the judgment methods of the problem phenomenon determination algorithm, then the problem phenomenon of the event information is determined to be the problem phenomenon corresponding to the problem phenomenon determination algorithm;

[0025] If the event information cannot satisfy the judgment method of each of the problem phenomenon determination algorithms, the event information will be pushed to the analyst.

[0026] Optionally, the method for determining whether the event information satisfies any of the problem phenomenon determination algorithms includes:

[0027] According to the problem time in the event information, a time series array of a first preset time interval of the problem time is obtained; the first preset time interval is a preset time before and a preset time after the problem time as the midpoint;

[0028] Extracting data features based on the current time series array, and judging whether the data features meet the current problem phenomenon to determine the judgment method of the algorithm;

[0029] If not, then, under the condition that the current preset time interval is within the preset maximum range, obtain the time series array of the next preset time interval of the problem time, and return to execute the step of extracting data features based on the current time series array, and judging whether the judgment method of the current problem phenomenon determination algorithm is satisfied according to the data features; wherein each preset time interval takes the problem time as the midpoint, and the minimum values ​​of adjacent preset time intervals differ by 1 preset time, and the maximum values ​​of adjacent preset time intervals differ by 1 preset time;

[0030] If so, it is determined that the event information satisfies the judgment method of the current problem phenomenon determination algorithm, and the preset time interval that satisfies the judgment method of the current problem phenomenon determination algorithm is used as the positioning time of the problem phenomenon.

[0031] Optionally, the event information corresponds to one or more problem phenomena;

[0032] When the event information corresponds to multiple problem phenomena, calling the analysis model corresponding to the problem phenomenon to analyze the event information, and obtaining the analysis result of the problem event includes:

[0033] The analysis models corresponding to the respective problem phenomena are called to analyze the event information respectively to obtain analysis results of the respective problem events.

[0034] The second aspect of the present application discloses an analysis system for assisted driving road testing, comprising:

[0035] An acquisition module is used to obtain event information of problem events that occur during the assisted driving test;

[0036] A preprocessing layer, used to determine the problem phenomenon corresponding to the event information according to the event information;

[0037] The analysis algorithm model pool is used to call the analysis model corresponding to the problem phenomenon to analyze the event information and obtain the analysis result of the problem event.

[0038] Optionally, also include:

[0039] An event class component module is used to construct an event class based on the event information; the event class includes a data part and a functional part, the data part is the event information; the functional part includes a video data request module, a timing data information request module, a storage module and a conclusion upload module; and the event class is transmitted to the preprocessing layer.

[0040] Optionally, the preprocessing layer is used to obtain required data according to the corresponding module of the functional part in the event class, and write the problem phenomenon into the data part of the event class;

[0041] The analysis algorithm model pool is used to obtain required data according to the corresponding modules of the functional part in the event class, and write the analysis results into the data part of the event class.

[0042] It can be seen from the above technical scheme that the present invention provides an analysis method for assisted driving road testing, including: obtaining event information of problem events occurring during the assisted driving test; determining the problem phenomenon corresponding to the event information based on the event information; calling the analysis model corresponding to the problem phenomenon to analyze the event information, and obtaining the analysis result of the problem event; that is, to analyze the problem events recorded in the road test, avoid manually analyzing each problem event with an analysis model, save the time cost of manual data analysis, and at the same time, before calling the analysis model, determine the problem phenomenon of the problem event, and then call the corresponding analysis model for targeted analysis to improve the accuracy of the analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] 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 described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0045] Figure 2 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0046] Figure 3 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0047] Figure 4 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0048] Figure 5 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0049] Figure 6 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0050] Figure 7 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0051] Figure 8 is a flow chart of an analysis method for assisted driving road testing provided by an embodiment of the present invention;

[0052] Fig. 9 It is a schematic diagram of an analysis system for assisted driving road testing provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0055] An embodiment of the present application provides an analysis method for assisted driving road testing, which is used to solve the problem in the prior art that all problems recorded by test engineers are transferred to analysts for data analysis on the same day. Since invalid problems account for a large proportion, this leads to a long analysis time, high labor costs, and limited value.

[0056] See also Figure 1 The analysis method of the assisted driving road test includes:

[0057] S101. Obtain event information of problem events that occurred during the assisted driving test.

[0058] Problem events: are actual problems caused by the assisted driving function encountered by road test engineers during actual road tests, such as mis-braking caused by misidentification of the perception algorithm, uncomfortable braking process, late braking, etc. When road test engineers discover these problems, they will record them in a table. The recorded content includes the event information of the problem event. The event information of the problem event includes: the time when the problem occurred, the description of the problem, and the preliminary classification of the problem, such as lateral control problems, perception problems, etc. The time when the problem occurred is accurate to seconds. The problem time can be the time when the road test engineer monitors the problem to occur. That is, when the road engineer detects the problem, he will record the corresponding occurrence time. The specific recorded content is as above.

[0059] That is to say, some problems may occur during the assisted driving process. One problem will be treated as a problem event, and the event information of each problem event will be recorded.

[0060] It should be noted that the event information of the problem event can be recorded by the road test engineer during the test process, and of course the automatic recording method is not excluded. Among them, the problem recording process is highly subjective and has a low degree of data dependence. It mainly relies on the actual vehicle performance of the assisted driving function in actual road conditions to record the problem.

[0061] The event information may include the problem time and problem description, etc. The problem description may include: the problem phenomenon and specific manifestations as considered by the test engineer, etc. Of course, the event information may also include information such as the vehicle type, and different model parameters may be matched based on the vehicle information.

[0062] Problem phenomena may refer to: false braking, false identification, etc.

[0063] Problem description: contains the problem phenomenon, but also has additional specific information, such as the current road scene, vehicle operating conditions, etc., so the problem description contains the problem phenomenon.

[0064] The problem phenomena recorded by road test engineers are usually superficial. They are not based on data summaries but on actual sensory records of road tests. There are errors, which is why the actual phenomenon of the problem needs to be recalculated through data features when allocating the model later.

[0065] S102: Determine the problem phenomenon corresponding to the event information according to the event information.

[0066] From the above description, it can be seen that the event information may include problem time and problem description, etc. Therefore, it is possible to obtain driving test data during the assisted driving process based on information such as problem time and problem description, and determine the problem phenomenon corresponding to the problem event based on the driving test data.

[0067] Specifically, driving test data within a preset range centered on the question time may be acquired based on the question time, and targeted driving test data may also be acquired based on the question description.

[0068] Of course, it is also possible to determine an algorithm based on the corresponding problem phenomenon and obtain the driving test data required by the problem phenomenon determination algorithm.

[0069] Then, the problem phenomenon analysis is performed based on the acquired driving test data to determine the problem phenomenon of the problem event. It should be noted that the problem phenomenon of the problem event is consistent with the problem phenomenon corresponding to the event information.

[0070] More specifically, corresponding data may be acquired according to the problem time to determine the corresponding problem phenomenon, and at the same time, after determining the problem phenomenon, the location time corresponding to the problem event may be determined.

[0071] It is worth noting that before step S102, an event class can also be constructed based on the event information; specifically, based on the event information, an event class is constructed. The event class includes a data part and a functional part, the data part is the event information; the functional part includes a video data request module, a timing data information request module, a storage module and a conclusion upload module.

[0072] There are four main functional modules in the event object, namely the event class, which are the time series data request module, the video data request module, the storage module and the conclusion upload module. Among them, the time series data request, video data request, storage and conclusion upload modules are customized functions developed based on time series database technology (such as IOTDB, etc.) and relational database (such as Mysql, etc.). The program is associated with the above databases through ORM (Object Relational Mapping), and the SQL Statement is executed in the program by connecting to the database engine, and finally the data in the database is transferred in the event object at the bottom of the program. These functions send data selection (SELECT), insertion (INSERT), update (UPDATE) and other requests to the database to obtain, upload or update data according to the SQLStatement automatically generated according to the input parameters, so as to realize data request, storage and upload.

[0073] The following is a description of each module:

[0074] Time series data request module: When executing the preprocessing layer, i.e. step 102, and analyzing the model, i.e. step 103, the time series array of the vehicle signal is needed. The data is stored in the database and needs to be queried through SQL (Structured Query Language). The time series data request module integrates this process. Specifically, the time series data request module generates a SQL statement based on the input; secondly, it queries the database for data; then, it resamples the data; and finally, it assigns the query result as an array to the event object (class). The resampling process is mainly because the fixed data sampling frequency recognized in the current system is 20ms / frame. However, in actual communication, there is a deviation in data sampling, that is, although the sampling frequency of most signals is also 20ms / s, the actual data per frame is sampled at 19.99x or 20.00x, which ultimately leads to a certain deviation in the length of the data. For example, the expected length of 10s data is 500 frames (20ms / frame), but the actual result is 50x frames or 49x frames, resulting in inconsistent data length. Therefore, the data length needs to be fixed through resampling.

[0075] Video data request module: In a specific analysis model, there is information that requires an artificial intelligence algorithm to obtain. The data format required by the algorithm is video format data. The function of this module is to obtain a video of a period of time before and after the event is sent, such as an MP4 format video, as the algorithm input.

[0076] Storage module: Stores event attributes, such as problem description, problem time, data used in the analysis process, and analysis conclusions, in the database.

[0077] Conclusion upload module: Upload the analysis conclusions to a platform visible to analysts, also implemented through SQL.

[0078] The event object, or event class, can be constructed in Python. This event object is a class in Python. The class not only has data information, but also can be given corresponding functions through class functions; for example:

[0079] ˋclass MyCase:

[0080] def__init__(self, number, description):

[0081] self.number = number

[0082] self.description=description

[0083] The above function defines a class, which contains two pieces of information: the first is number and the second is description, which also constitutes a piece of data information. At the same time, this class only has information and no function.

[0084] For example, we need a function to solve the quadratic number in this class:

[0085] ˋclass MyCase:

[0086] def__init__(self, number, description):

[0087] self.number = number

[0088] self.description=description

[0089] defpow(self):

[0090] self.number = self.number**2

[0091] The functions in the event object are assigned in this way. The purpose of this is that after the event is passed into the analysis model as input, the analysis model needs to calculate the relevant information through the data. At this time, the time series data request module is needed to assign the specified data to the event object, so that it can be used as needed and save memory space expenses.

[0092] Variables in Python are all objects, such as event classes, event structures, etc.

[0093] That is to say, in step S102 and step S103, corresponding modules may be called to obtain data or perform actions such as storage and uploading.

[0094] S103: Call the analysis model corresponding to the problem phenomenon to analyze the event information and obtain the analysis result of the problem event.

[0095] It should be noted that different problem phenomena correspond to different analysis algorithm model pools; therefore, it is necessary to call the analysis model corresponding to the problem phenomenon to analyze the current event information and obtain the analysis result of the problem event.

[0096] The specific analysis process may be that after calling the corresponding analysis model, the analysis algorithm model pool obtains the corresponding data according to the event information for analysis to obtain the analysis result.

[0097] It can be known from the above description that the positioning time can also be obtained in step S102, so the data corresponding to the positioning time can be obtained for analysis to obtain the analysis result.

[0098] Different analysis models may have different analysis logics. Here, the analysis process of each analysis algorithm model pool is described. It should be noted that different analysis models can obtain the required data for analysis based on event information.

[0099] That is, before the problem is released to the analyst, the processed analysis results are provided to the analysis engineer in the form of text, statistical data, etc. through the designed algorithm logic.

[0100] In this embodiment, event information of problem events occurring during the assisted driving test is obtained; based on the event information, the problem phenomenon corresponding to the event information is determined; the analysis model corresponding to the problem phenomenon is called to analyze the event information to obtain the analysis result of the problem event; that is, the problem events recorded in the road test are analyzed to avoid manual analysis of each problem event using the analysis model, thus saving the time cost of manual data analysis. At the same time, before calling the analysis model, the problem phenomenon of the problem event is determined, and then the corresponding analysis model is called for targeted analysis to improve the accuracy of the analysis.

[0101] It should be noted that in the division of labor analysis method, due to the inconsistent ability levels of analysts, there may be different conclusions for the same problem, and incorrect analysis conclusions. In addition, the current analysis method for road test problems has inherent analysis ideas and processes. After receiving the problem, the analyst will first check the data according to the inherent analysis ideas and prioritize the elimination of the problem.

[0102] In this embodiment, an analysis model is used for analysis, which avoids the problem of using artificial standards, inconsistent levels and improves the accuracy of analysis. At the same time, especially when dealing with large-scale, repetitive and unavoidable analysis problems, the program algorithm has an absolute advantage in execution efficiency and accuracy, which reduces the repeated use of manpower and improves analysis efficiency.

[0103] Optionally, the specific process of determining the problem phenomenon corresponding to the event information according to the event information in step S102 may be:

[0104] First, various problem phenomenon determination algorithms are used to traverse the event information; secondly, it is determined whether the event information meets the judgment method of each problem phenomenon determination algorithm; if the event information meets the judgment method of any problem phenomenon determination algorithm, the problem phenomenon of the event information is determined to be the problem phenomenon corresponding to the problem phenomenon determination algorithm; if the event information does not meet the judgment method of each problem phenomenon determination algorithm, the event information is pushed to the analyst.

[0105] This step obtains the specified data through the timing data request module in the event object, such as vehicle body signal data (vehicle speed, acceleration, steering wheel angle, etc.) or algorithm status signal, and makes a judgment based on the logic of the known problem phenomenon, outputs the actual problem phenomenon of the problem event, and outputs the location time of the problem phenomenon.

[0106] The location time of the problem here is the specific location array index of the time when the problem occurred, accurate to the frame. For example, in the actual process of road testing, the road test engineer recorded problem A at xx minutes and xx seconds. The time of this problem is only accurate to seconds. Since the sampling frequency of the data is 20ms / frame, the time provided by the road test engineer deviates from the actual time when the problem occurred. This deviation is not accepted in the program. When the event object passes through the preprocessing layer, it will determine whether a problem has occurred based on specific logic. The time interval or moment that meets the condition for the problem to occur is the location time of the problem phenomenon accurate to the frame. This location time may be a moment or a time interval.

[0107] Taking the problem of double follow-up stop as an example, the road test engineer recorded a double follow-up stop problem at 10:43:20 on December 12, 2023. After this problem was constructed into an event object, it was first passed to this step. The execution logic of this step is:

[0108] First, the problem events are input into all existing problem phenomenon determination algorithms in parallel, such as the secondary follow-up stop, deceleration control abnormality and other problem phenomenon determination algorithms; secondly, each problem phenomenon judgment algorithm obtains the timing array through the timing signal request module according to the data required in the algorithm; then, the algorithm extracts the data features and determines whether it belongs to this type of problem based on the data features; finally, if the judgment is true, the true value of the problem phenomenon and the positioning time are output, otherwise the attributes in the event class are not changed.

[0109] Optionally, a method for determining whether event information satisfies any problem phenomenon determination algorithm includes:

[0110] According to the problem time in the event information, a time series array of a first preset time interval of the problem time is obtained; the first preset time interval is a preset time before and a preset time after the problem time as the midpoint.

[0111] Specifically, the first preset time interval is [t1, t2], where the problem time minus t1 equals 1 preset time, and t2 minus the problem time equals 1 preset time. The preset time can be 10s, and of course it can also be other values. We will not go into details here, and it depends on the actual situation. They are all within the protection scope of this application.

[0112] The data features are extracted based on the current time series array, and the judgment method of the algorithm is determined based on whether the data features meet the current problem phenomenon.

[0113] If not, then while ensuring that the current preset time interval is within the preset maximum range, obtain the time series array of the next preset time interval of the problem time, and return to execute the step of extracting data features based on the current time series array, and judging whether the judgment method of the algorithm for determining the current problem phenomenon is satisfied based on the data features; wherein each preset time interval takes the problem time as the midpoint, and the minimum values ​​of adjacent preset time intervals differ by 1 preset time, and the maximum values ​​of adjacent preset time intervals differ by 1 preset time.

[0114] To ensure that the current preset time interval is within the preset maximum range, the number of times the timing array is obtained can be repeatedly executed, or the maximum time range can be set. I will not go into details here, and it depends on the actual situation. All of them are within the protection scope of this application.

[0115] Assume that the i-th preset time interval is [t i，1 , t i，2 ], the i-1th preset time interval [t i-1，1 , t i-1，2 ], the i-1th preset time interval [t i+1，1 , t i+1，2 ], where t i+1，1 -t i，1 = preset time, t i，1 -t i-1，1 = preset time, t i+2，1 -t i，2 = preset time, t i，2 -t i-1，2 =Preset time.

[0116] If so, it is determined that the event information satisfies the judgment method of the current problem phenomenon determination algorithm, and the preset time interval that satisfies the judgment method of the current problem phenomenon determination algorithm is used as the location time of the problem phenomenon.

[0117] The following is an example of the secondary stop judgment algorithm:

[0118] First, the time series signal request module in the event object is used to obtain the time series array (a[10, 10]) of 10 seconds before and after 10 seconds of 10:43:20:0000 (T1). The sampling frequency is 20ms / frame, so each number in the array represents the vehicle speed signal of the current frame, and the same is true for other signals. 10 seconds is the preset time.

[0119] Secondly, the secondary follow-up stop judgment algorithm uses specific logic to determine whether the problem event is a secondary follow-up stop event.

[0120] a) The first time the vehicle speed is equal to 0 within the time interval of 10 seconds before and after 10 seconds is recorded as t1. If this time does not exist, the timing signal request module is re-executed to obtain the vehicle speed signal within 20 seconds before and after the time recorded by the road test engineer (a[20, 20]) and perform the judgment again. Of course, the number of loops can be configured, that is, there is a loop limit in this process. If t1 is not found after more than n1 times, it will not be considered as a secondary follow-up stop problem.

[0121] b) Find the moment t2 when the vehicle speed is greater than 0 for the first time after moment t1. If t2 does not exist, re-execute the timing signal request module to obtain the vehicle speed signal within 10 seconds before and after the moment recorded by the road test engineer and perform the judgment again. This process has a loop limit. If t2 is still not found after more than n2 times, it will not be considered as a secondary follow-up stop problem.

[0122] c) After time t2, if the vehicle's starting mileage is less than x1 meters or the starting time is less than x2 milliseconds and the vehicle stops again, that is, the vehicle speed is equal to 0 again, then the judgment condition of the secondary stop problem phenomenon is met, and the output is true. A secondary stop attribute will be added to the problem phenomenon field of the problem event, and [t1, t2] is the interval of the location time of the problem, that is, the problem occurs between the t1th frame and the t2th frame in the interval a[10*n1, 10*n2]. Otherwise, it is false and will not affect the attributes of the event class.

[0123] d) Carry out other problem phenomenon judgment algorithms.

[0124] It should be noted that the various problem phenomenon judgment algorithms can be executed in parallel or in sequence, which will not be described here one by one, depending on the actual situation.

[0125] e) Jump out of the preprocessing layer.

[0126] After all problem phenomenon judgment algorithms are finished, the pre-processing layer is exited, that is, step S102 is ended.

[0127] It should be noted that after the event object jumps out of the preprocessing layer, there are two possibilities. The first case is that the problem phenomenon is found. In this case, it will be indexed into the corresponding analysis model according to the problem phenomenon. For example, through the calculation of the vehicle speed and other signal characteristics of event A, it is found that there are problems with deceleration control abnormality and braking abnormality. The event will be indexed by the program into the two models of deceleration control abnormality and braking abnormality for analysis. Event B only calculates the problem of braking abnormality, and the event will only be indexed into the braking abnormality analysis model; among them, the problem phenomenon is determined based on the accumulated experience of road tests. There are currently more than 10 categories, and coverage can be continuously developed in the subsequent accumulation process.

[0128] The second situation is that the problem phenomenon field is empty, that is, there is no corresponding problem phenomenon judgment algorithm. The system believes that there is no corresponding analysis model for the phenomenon. This type of problem may be an abnormal road test situation that has not appeared in the system. Similar events will not be indexed into any model and will be submitted to analysts for targeted analysis. After the manual analysis draws a conclusion, iterative development will be carried out based on the new problem phenomenon and the corresponding analysis model to cover subsequent similar problem events.

[0129] Optionally, the specific process of the above step S103, calling the analysis model corresponding to the problem to analyze the event information and obtaining the analysis result of the problem event may be:

[0130] First, the analysis model corresponding to the problem phenomenon is used as the target analysis model; secondly, the target analysis model is called, and the vehicle model parameters corresponding to the event information are input into the target analysis model; then, the target analysis model performs analysis and processing to obtain the analysis results of the problem event.

[0131] That is to say, multiple analysis models are configured in the analysis algorithm model pool, and the corresponding analysis algorithm model pool is called in a targeted manner according to the problem phenomena of the problem event to conduct targeted analysis and improve the accuracy of the analysis.

[0132] Optionally, the event information corresponds to one or more problem phenomena.

[0133] When there are multiple problem phenomena corresponding to the event information, the specific process of step S103, calling the analysis model corresponding to the problem to analyze the event information and obtaining the analysis results of the problem events, can be: calling the analysis models corresponding to each problem phenomenon to analyze the event information separately and obtain the analysis results of each problem event.

[0134] That is to say, multiple analysis models can be used to analyze the same problem event, and whether the main problem event involves multiple problem phenomena. When using multiple analysis models for analysis, the analysis results of each analysis model can be used as the analysis result of the problem event.

[0135] It should be noted that the parameters required for different vehicle models are not the same. Therefore, the corresponding parameters can be matched according to the vehicle model. For example, the current vehicle model is obtained based on the vehicle model data segment in the event information, and the required parameters of the target analysis model for the current vehicle model are obtained from the algorithm input parameter document and input into the target analysis model; then the analysis model analyzes the problem event based on the vehicle model parameters.

[0136] Specifically, the algorithm input parameter document stores the parameters required by different analysis models for different vehicle models, and the analysis model then obtains the corresponding parameters based on the vehicle model. In other words, the algorithm input document is the relevant parameters of the assisted driving function, which is used to ensure that the system can be adapted to different vehicle models by modifying parameters, decoupling the correlation between the automatic analysis system and the vehicle model, so that the analysis logic can be reused in a large number of different vehicle models.

[0137] In this embodiment, when the assisted driving algorithm is adapted to different OEM (Original Equipment Manufacturer) models, the calibration of the algorithm parameters will be adjusted according to the actual road driving experience, customer needs and other situations. The adjustment of these calibration quantities is the responsibility of the algorithm engineer, and the output of the phased adjustment or the final adjustment is finally identified. Therefore, the system automatically identifies these parameters by reading the relevant parameter files, and no manual modification is required. The calibration quantity is a constant.

[0138] Optionally, the analysis model performs analysis and processing to obtain analysis results of the problem event, including:

[0139] First, the driving test data corresponding to the problem phenomenon within the positioning time corresponding to the problem phenomenon is obtained; then, according to the driving test data, logical judgment is performed on each parameter in the driving test data in turn according to the structure of the link tree to obtain the analysis result.

[0140] The positioning time may be obtained in step S102, and further, accurate data information may be found through the positioning time, thereby improving the analysis accuracy and efficiency.

[0141] Specifically, data features may be proposed based on the driving test data, and a link tree may be constructed based on the data features, and the link tree may be executed to obtain analysis results.

[0142] Optionally, based on the driving test data, logical judgment is performed on each parameter in the driving test data in sequence according to the structure of the link tree to obtain analysis results, including:

[0143] Firstly, according to the driving test data, logical judgment is made on each parameter in the driving test data in turn according to the structure of the link tree to determine the analysis conclusion; secondly, the corresponding solution and problem type are matched according to the analysis conclusion; then, the analysis conclusion, solution and problem type are taken together as the analysis result.

[0144] After determining the analysis conclusion, the corresponding solution and problem type can be determined according to the corresponding relationship between the analysis conclusion, the solution and the problem type. The corresponding relationship can be pre-configured and updated according to the actual situation. It is not specifically limited here and can be determined according to the actual situation, all within the scope of protection of this application.

[0145] After analyzing the problem event, a variety of information can be obtained, such as analysis conclusions, problem types, solutions, and data accumulation. Among them, the analysis conclusions and problem solutions are equivalent to the conclusions of manual analysis. Problem types and data accumulation are additional outputs provided by the system. The main function is data accumulation, which will be used in the future for statistics of common problems in road testing, problem data mining, and machine learning of problem events.

[0146] That is to say, the internal execution logic of the analysis model after accessing the specific event object is explained by taking the secondary follow-up stop event as an example. After step S102, the event object adds new attributes, such as the problem phenomenon and the positioning time. Finally, the problem event is indexed to the secondary follow-up stop analysis model through the problem phenomenon secondary follow-up stop. After the secondary follow-up stop analysis model is accessed with parameters, information is extracted according to the developed logic, a link tree node LinkedTreeNode is constructed, and the link tree LinkedTree is executed. The details of this process are as follows:

[0147] 1. Obtain the parameters required by the secondary follow-up stop analysis model from the algorithm input parameter document. For example, the secondary follow-up stop analysis model relies on the following stop distance m1 that the algorithm expects the assisted driving function to follow the vehicle in front.

[0148] Specifically, the algorithm parameters are obtained: the starting relative distance m1 expected by the algorithm, where m1 is the relative distance to the preceding vehicle.

[0149] Assisted driving is the result of the integration of multiple algorithms. These algorithms include: perception algorithms, decision algorithms, control algorithms, etc. These algorithms determine the driving mode and driving path of the vehicle, and send the driving mode and path information of the vehicle to the vehicle's actuators through the controller for execution, ultimately implementing the function.

[0150] Faults may exist in any one or more of the above processes. For example, a sensing fault may cause the vehicle to misidentify or miss the vehicle in front. Or there may be a delay in the actuator executing the driving mode information sent by the controller, resulting in untimely execution.

[0151] The algorithm referred to here refers to the integration of the above multiple algorithms, that is, the assisted driving algorithm. Therefore, the parameters obtained are the thresholds that the vehicle is expected to reach when the assisted driving algorithm is executed by the vehicle. For example, when the algorithm expects the vehicle to follow the stop action, the final distance to the vehicle in front is maintained.

[0152] The problem event refers to the following stop distance m1 that the algorithm expects, but the final following stop distance is d1, which is greater than m1. This does not meet the algorithm's expectations, and a failure occurs.

[0153] The purpose of this system is to determine the specific cause of the fault through the vehicle driving data for a period of time before and after the fault occurs. This process is performed by a problem analysis model, such as a secondary follow-up stop analysis model.

[0154] When analyzing the model to determine the specific cause of a fault, we occasionally rely on the specific algorithm logic of the assisted driving algorithm. We mainly troubleshoot the problem through objective data performance judgment, find the specific data that causes the problem, give the customer an explanation, and save analysis time for algorithm engineers.

[0155] Troubleshoot problems through objective data performance judgment: For example, when the following stop distance is too far, first determine whether the deceleration request sent by the assisted driving algorithm to the actuator is too large, causing the vehicle to brake too hard; if so, that is, the assisted driving sends a large deceleration to the actuator through the controller, that is, the deceleration sent by the algorithm is unreasonable, and the algorithm module needs to be readjusted. The problem of following the stop distance is caused by an algorithm problem; if not, that is, the deceleration request sent by the assisted driving algorithm is appropriate, but the actuator overshoots when executing the deceleration request sent by the controller (overshoot refers to the algorithm sending a 1m / s 2 The deceleration of the vehicle is 2m / s. 2 , which is twice the expected deceleration), then the problem is caused by excessive actuator response and is an actuator problem.

[0156] In the above, when performing LinkedTree judgment of the analysis model, the assisted driving algorithm was not used to obtain information. The cause of the problem was discovered only through the objective data performance of the vehicle, and the analysis of the problem was automatically completed.

[0157] 2. Obtain the time series data required by the secondary follow-up and stop analysis model through the time series data request module in the event object.

[0158] For example, the secondary stop-and-follow analysis model relies on the time series data of several signals, such as the vehicle speed, the relative distance to the target vehicle, and the relative speed to the target vehicle. Different analysis models rely on different data, which mainly depends on the analysis logic.

[0159] More specifically, the data obtained is as follows:

[0160] Value: The time t3 when the algorithm requests deceleration; that is, the algorithm starts to send deceleration requests to the vehicle actuator through the controller, and the algorithm expects the vehicle to start decelerating. The relative distance from the target vehicle at the time of following and stopping is d1; that is, the relative distance from the target vehicle when the vehicle speed is equal to zero for the first time during the process of following and stopping the vehicle in front.

[0161] Array: vehicle speed Vx[t1, t2]; vehicle deceleration Ax[t1, t2]; relative distance to the target vehicle RelativeDistanceX[t1, t2]; target vehicle speed CIPVVX[t1, t2]; algorithm request deceleration state RequestAxSts; algorithm request deceleration value RequestAxValue[t3, t1].

[0162] Among them, Vx[t1, t2] is the actual driving speed of the ego vehicle from time t1 to time t2; Ax[t1, t2] is the acceleration and deceleration of the ego vehicle from time t1 to time t2; RelativeDistanceX[t1, t2] is the longitudinal distance between the ego vehicle and the target vehicle from time t1 to time t2; Target vehicle speed CIPVVX[t1, t2]: the speed of the target vehicle from time t1 to time t2; Algorithm request deceleration state RequestAxSts: the state of the algorithm's expected ego vehicle deceleration, 0 means the algorithm does not expect the ego vehicle to decelerate, 1 means the algorithm starts to issue a deceleration request to the vehicle, that is, the algorithm expects the vehicle speed to decrease. Algorithm request deceleration value RequestAxValue[t3, t1] is the deceleration request value issued by the algorithm from the moment the algorithm starts to expect the vehicle to decelerate to the moment the ego vehicle starts.

[0163] Get logical information through algorithms:

[0164] Node cond1: RelativeDistanceX[t1]>m1.

[0165] Node cond2: Under the premise that the two vehicles are stationary within [t1, t2], RelativeDistanceX[t2]>m1.

[0166] Node cond3: At any moment, the deceleration of the ego vehicle does not reach the deceleration expected by the algorithm as required. For example, when the algorithm issues a request deceleration of 10m / s2, the acceleration of the ego vehicle reaches this value as expected, and then the algorithm starts to cancel the deceleration, but the real vehicle does not respond in time during this process, resulting in the deceleration not being canceled as expected.

[0167] Node cond4: CIPVVX[t1, t2] has a value greater than 0 and the process causes RelativeDistanceX>m1.

[0168] 3. Make logical judgments based on the structure of the LinkedTree.

[0169] Specifically, a LinkedTreeNode is constructed based on the above nodes, such as Figure 5 .

[0170] There are one or more logical judgments in each node of the tree, and the final output is the judgment result and the pointers to the next two nodes. This logic tree will continue until the next node is empty, that is, it reaches a conclusion node. The final execution result will connect all the traversed nodes in series and eventually become a linked list structure. The conclusions of all nodes in this linked list are added together to form the final output result, which is composed of process nodes and conclusion nodes.

[0171] like Figure 6 , Figure 7 and Figure 8 As shown in the figure, the result of adding the process nodes and the conclusion nodes is the analysis conclusion, problem type and solution to the problem that are finally output to the analyst. At the same time, after the execution is completed, the event and the corresponding output will be stored in the database for other possible subsequent applications, such as accumulating test cases, data mining, etc.

[0172] like Figure 6 As shown, the process nodes are: Node cond1: True, the first stop distance d1 is greater than the algorithm expected value m1, and there is a problem of stopping too far; Node cond3: True, the actual vehicle actuator responds to the algorithm to cancel the request for deceleration with a delay of S1 milliseconds during the stop process. Conclusion node, analysis conclusion: The first stop distance d1 is greater than the algorithm expected value m1, and there is a problem of stopping too far; during the stop process, the actual vehicle actuator responds to the algorithm to cancel the request for deceleration with a delay of S1 milliseconds, causing the vehicle to brake in advance. At this time, the stop distance reaches the threshold of the algorithm request to start, and the vehicle starts again. This problem is caused by the slow response of the actual vehicle; Solution: Adjust the algorithm constant threshold or feedback to the OEM; Problem type: Algorithm problem, OEM problem.

[0173] like Figure 7As shown, process nodes: Node cond1: False, the first follow-stop distance is d1, which meets the algorithm expectation value m1; Node cond2: False, None; Node cond4: True, the relative distance of the front vehicle starting during parking is larger than m1. Conclusion node, analysis conclusion: The first follow-stop distance is d1, which meets the algorithm expectation value m1, and the relative distance of the front vehicle starting during parking is larger than m1. The start-stop caused by non-self-vehicle function factors is not a problem. Solution: None; Problem type: None.

[0174] like Figure 8 As shown, process nodes: Node cond1: False, the first stop distance is d1, which meets the algorithm's expected value m1; Node cond2: False, None; Node cond4: False, None. Conclusion node, analysis conclusion: The first stop distance is d1, which meets the algorithm's expected value m1. There is no other analysis logic for this problem, and the program cannot be analyzed for the time being. Solution: None; Problem type: None.

[0175] 4. Get the analysis conclusion.

[0176] Specifically, execute the linked tree LinkedTree to obtain the analysis conclusion.

[0177] 5. Upload the conclusion.

[0178] 6. Complete the analysis.

[0179] Another embodiment of the present application provides an analysis system for assisted driving road testing.

[0180] See also Fig. 9 , the analysis system for assisted driving road testing includes:

[0181] The acquisition module is used to obtain event information of problem events that occur during the assisted driving test.

[0182] The preprocessing layer is used to determine the problem phenomenon corresponding to the event information based on the event information.

[0183] The analysis algorithm model pool is used to call the analysis model corresponding to the problem phenomenon to analyze the event information and obtain the analysis results of the problem event.

[0184] The analysis model of each problem phenomenon is a class composed of Python code. This class is mainly instantiated by introducing an event object, and then the functions and methods in the data acquisition module are used according to the functions in the event object. The information that is beneficial to the analysis process will be calculated through these data, and finally the data structure of the linked tree node LinkedTreeNode (such as Figure 5 )Connect scattered information to form analysis process, conclusions and solutions.

[0185] These data can be time series data or video data; time series data, for example, the vehicle body signals needed to analyze the misbraking problem, such as vehicle speed, yaw angle, acceleration and deceleration, etc. Algorithm signals: acceleration and deceleration requested by the algorithm, that is, arrays.

[0186] In this embodiment, compared with manually analyzing all problems in this way, the system greatly reduces the time cost of analysis work, and also reduces the analysis deviation and errors that may be caused by manual analysis. In terms of different vehicle models, the system can be widely adapted by modifying the threshold in the algorithm input document. Therefore, the application of the system is not limited by the vehicle model and has a wide range of applications. In response to the update and iteration of the assisted driving function algorithm, the system has greatly sorted out and improved the complex conditional statement logic through a binary tree structure. It only needs to fine-tune the specified nodes that have changed to adapt to the update of the assisted driving function algorithm, without the need to make large-scale changes to a certain analysis model, which will bring additional development costs. Therefore, the structure of the system is flexible and stable.

[0187] Optionally, the analysis system for assisted driving road testing also includes:

[0188] The event class component module is used to construct an event class based on event information; the event class includes a data part and a functional part, the data part is the event information; the functional part includes a video data request module, a time series data information request module, a storage module and a conclusion upload module; and the event class is transmitted to the preprocessing layer.

[0189] Optionally, the preprocessing layer is used to obtain required data according to the corresponding modules of the functional part in the event class, and write the problem phenomenon into the data part of the event class.

[0190] There are some pre-written logic algorithms in the pre-processing layer, such as: mis-braking judgment algorithm, lateral control abnormality judgment algorithm, longitudinal control abnormality judgment algorithm, braking process abnormality judgment algorithm, and secondary follow-up stop judgment algorithm; after the event object is passed into the pre-processing layer as a parameter, these logics will be executed accordingly to judge the problem phenomenon of the current event. The problem phenomenon summarized by the above logic algorithm is the final problem phenomenon.

[0191] The analysis algorithm model pool is used to obtain the required data according to the corresponding modules of the functional part in the event class, and write the analysis results into the data part of the event class.

[0192] This application mainly replaces manual problem analysis after the data upload process, and directly obtains problem analysis conclusions and other related information through the analysis system, such as Figure 2As shown, problem events and algorithm input documents are the input of the analysis system, and analysis conclusions, problem types, solutions and data accumulation are the outputs.

[0193] It should be noted that if Figure 3 As shown in the figure, the internal execution logic diagram of the automated analysis system. After receiving the incoming event, the program integrates the event information first. For example, the time when the problem occurred, the problem description, etc., are constructed as objects in the program, which are subsequently called event objects. The purpose of each module in the event object is to provide corresponding functional methods so that it can obtain the necessary data and information and realize the necessary functions through these methods in the analysis model. These modules are subsidiary functions of the event object, which are independent of each other and can be executed through program calls when needed to achieve the purpose. After the event object is passed to the preprocessing layer, the preprocessing layer calculates the problem phenomenon that occurs within a specific time range based on the data characteristics of the event object, and finally indexes it into the corresponding analysis model based on the problem phenomenon, and analyzes the problem event in combination with the algorithm input document to draw a conclusion.

[0194] After the problem event is input into the analysis system and calculated by the corresponding analysis model, it is output as four types of information, namely analysis conclusion, problem type, solution and data accumulation. Among them, the analysis conclusion and the solution to the problem are equivalent to the conclusion of manual analysis. Problem type and data accumulation are additional outputs provided by the system. Their main function is data accumulation, which will be used in the future for statistics of common problems in road testing, problem data mining, machine learning of problem events, etc.

[0195] like Figure 4 As shown, the analysis model extracts information based on the event object and algorithm input parameter documents. For example, the artificial intelligence algorithm class extracts driving scene information, automatic driving target information, etc.; the automatic driving function algorithm extracts automatic body state information, algorithm function state information, driving function state information and perception algorithm information, etc.; the basic algorithm extracts the difference information between the algorithm request value and the actual response value (steering wheel angle request, acceleration and deceleration request), target vehicle recognition abnormality judgment information, lane recognition abnormality judgment information, function abnormality judgment information, body abnormality judgment information and OEM parts abnormality judgment information, etc. It should be noted that the artificial intelligence algorithm class, automatic driving function algorithm and basic algorithm are the algorithms integrated by the vehicle itself, and then the information corresponding to the algorithm integrated by the vehicle itself is extracted.

[0196] In this embodiment, the analysis system uses the problem events of road tests as the core data structure, and develops peripheral algorithms, core functions, and analysis models based on this data structure. Ultimately, under the interaction of various development modules, the data structure can automatically index the analysis model based on the problem phenomena shown by the current data, analyze the problem, and finally store the results and submit them to the analysts.

[0197] The features recorded in the various embodiments in this specification can be replaced or combined with each other, and the same and similar parts between the various embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.

[0198] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0199] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An analysis method for assisted driving road testing, characterized in that: include: Obtain event information of problem events that occurred during assisted driving testing; Determine, according to the event information, the problem phenomenon corresponding to the event information; The analysis model corresponding to the problem phenomenon is called to analyze the event information to obtain the analysis result of the problem event.

2. The analysis method for assisted driving road test according to claim 1, characterized in that: The calling of the analysis model corresponding to the problem to analyze the event information to obtain the analysis result of the problem event includes: Taking the analysis model corresponding to the problem phenomenon as the target analysis model; Calling the target analysis model, and inputting the vehicle model parameters corresponding to the event information into the target analysis model; The target analysis model performs analysis processing to obtain an analysis result of the problem event.

3. The analysis method for assisted driving road test according to claim 2, characterized in that: The analysis model performs analysis and processing to obtain analysis results of the problem event, including: Acquire driving test data corresponding to the problem phenomenon within the positioning time corresponding to the problem phenomenon; Based on the driving test data, logical judgment is performed on each parameter in the driving test data in sequence according to the structure of the link tree to obtain an analysis result.

4. The analysis method for assisted driving road test according to claim 3, characterized in that: According to the driving test data, logical judgment is performed on each parameter in the driving test data in sequence according to the structure of the link tree to obtain analysis results, including: Based on the driving test data, logical judgment is performed on each parameter in the driving test data in sequence according to the structure of the link tree to determine the analysis conclusion; Matching corresponding solutions and problem types according to the analysis conclusions; The analysis conclusion, solution and problem type are collectively referred to as the analysis result.

5. The analysis method for assisted driving road test according to claim 1, characterized in that: The determining, according to the event information, a problem phenomenon corresponding to the event information includes: The event information is traversed using various problem phenomenon determination algorithms; A method for determining whether the event information satisfies each of the problem phenomenon determination algorithms; If the event information satisfies any of the judgment methods of the problem phenomenon determination algorithm, then the problem phenomenon of the event information is determined to be the problem phenomenon corresponding to the problem phenomenon determination algorithm; If the event information cannot satisfy the judgment method of each of the problem phenomenon determination algorithms, the event information will be pushed to the analyst.

6. The analysis method for assisted driving road test according to claim 5, characterized in that: The method for determining whether the event information satisfies any of the problem phenomenon determination algorithms includes: According to the problem time in the event information, a time series array of a first preset time interval of the problem time is obtained; the first preset time interval is a preset time before and a preset time after the problem time as the midpoint; Extracting data features based on the current time series array, and judging whether the data features meet the current problem phenomenon to determine the judgment method of the algorithm; If not, then, under the condition that the current preset time interval is within the preset maximum range, obtain the time series array of the next preset time interval of the problem time, and return to execute the step of extracting data features based on the current time series array, and judging whether the judgment method of the current problem phenomenon determination algorithm is satisfied according to the data features; wherein each preset time interval takes the problem time as the midpoint, and the minimum values ​​of adjacent preset time intervals differ by 1 preset time, and the maximum values ​​of adjacent preset time intervals differ by 1 preset time; If so, it is determined that the event information satisfies the judgment method of the current problem phenomenon determination algorithm, and the preset time interval that satisfies the judgment method of the current problem phenomenon determination algorithm is used as the positioning time of the problem phenomenon.

7. The analysis method for assisted driving road test according to claim 1, characterized in that: The event information corresponds to one or more problem phenomena; When the event information corresponds to multiple problem phenomena, calling the analysis model corresponding to the problem phenomenon to analyze the event information, and obtaining the analysis result of the problem event includes: The analysis models corresponding to the respective problem phenomena are called to analyze the event information respectively to obtain analysis results of the respective problem events.

8. An analysis system for assisted driving road testing, characterized in that: include: An acquisition module is used to obtain event information of problem events that occur during the assisted driving test; A preprocessing layer, used to determine the problem phenomenon corresponding to the event information according to the event information; The analysis algorithm model pool is used to call the analysis model corresponding to the problem phenomenon to analyze the event information and obtain the analysis result of the problem event.

9. The analysis system for assisted driving road testing according to claim 8, characterized in that: Also includes: An event class component module is used to construct an event class according to the event information; The event class includes a data part and a functional part, the data part is the event information; the functional part includes a video data request module, a timing data information request module, a storage module and a conclusion upload module; and the event class is transmitted to the preprocessing layer.

10. The analysis system for assisted driving road testing according to claim 9, characterized in that: The pre-processing layer is used to obtain required data according to the corresponding module of the functional part in the event class, and write the problem phenomenon into the data part of the event class; The analysis algorithm model pool is used to obtain required data according to the corresponding modules of the functional part in the event class, and write the analysis results into the data part of the event class.

Images