CN115526263A

CN115526263A - Vehicle abnormity positioning method, device, equipment and storage medium

Info

Publication number: CN115526263A
Application number: CN202211257618.7A
Authority: CN
Inventors: 莫国龙; 顾瑞红
Original assignee: Ecarx Hubei Tech Co Ltd
Current assignee: Ecarx Hubei Tech Co Ltd
Priority date: 2022-10-14
Filing date: 2022-10-14
Publication date: 2022-12-27

Abstract

The embodiment of the invention provides a method, a device, equipment and a storage medium for positioning the abnormity of a vehicle, wherein the method comprises the following steps: determining target combination data corresponding to the vehicle when the function operation is abnormal according to at least one type of operation data generated in the process of performing the function operation on the vehicle, wherein the target combination data comprise operation problem types and characteristic data associated with the operation problem types; training a set machine learning model according to the target combination data to obtain a classification model; and positioning an abnormal operation object of the vehicle when the function operation is abnormal according to the classification model. By utilizing the method, problem data with abnormal function operation is determined through each operation data, then model training is carried out based on the problem data to obtain a classification model, and an operation abnormal object influencing the decision effect under different problems is determined according to the classification model, so that the automatic positioning of the operation abnormal object is realized, the efficiency and the accuracy of abnormal positioning are improved, and the labor cost is reduced.

Description

Vehicle abnormity positioning method, device, equipment and storage medium

Technical Field

The invention relates to the technical field of vehicle abnormity positioning, in particular to a vehicle abnormity positioning method, device, equipment and storage medium.

Background

In the research and development of the intelligent vehicle, various operations are required, and more related data can be acquired, calculated and generated through a sensor and a software algorithm in the operation process. The acquired data is mainly used for software decision, and after calculation of each different decision module is completed, a decision result can be generated. The decision data acquired by the sensors or calculated and generated by different software play a key role in the decision of the whole intelligent driving, so that the quality of each data can be detected, the influence degree of various data on the whole intelligent driving can be determined, and the decision-making method is an important basis for determining the automatic driving capability.

With the expansion of the software scale and the distributed characteristic, the incidence rate of the defects of the algorithm decision module is higher and higher. In the prior art, the defect positioning mode of the algorithm decision module is mainly to detect and identify related data manually based on historical experience, and due to the fact that the data size is large, software defects are difficult to probe, rapid positioning is difficult, and labor cost is wasted.

Disclosure of Invention

The embodiment of the invention provides a method, a device, equipment and a storage medium for positioning an abnormal vehicle, which are used for quickly and automatically positioning an abnormal running object, improving the efficiency and accuracy of abnormal positioning and reducing the labor cost.

In a first aspect, the present embodiment provides a method for locating an abnormality of a vehicle, the method including:

determining target combination data corresponding to the vehicle when the vehicle is abnormal in functional operation according to at least one type of operation data generated in the process of performing functional operation on the vehicle, wherein the target combination data comprise operation problem types and characteristic data associated with the operation problem types;

training a set machine learning model according to the target combination data to obtain a classification model;

and positioning an abnormal operation object of the vehicle when the function operation is abnormal according to the classification model.

In a second aspect, the present embodiment provides an abnormality locating device for a vehicle, the device including:

the data determining module is used for determining target combined data corresponding to the vehicle when the vehicle is abnormal in functional operation according to at least one type of operation data generated in the process of performing functional operation on the vehicle, wherein the target combined data comprises operation problem types and characteristic data associated with the operation problem types;

the model determining module is used for training a set machine learning model according to the target combination data to obtain a classification model;

and the abnormity positioning module is used for positioning an abnormal operation object of the vehicle when the function operation is abnormal according to the classification model.

In a third aspect, this embodiment provides an electronic device, including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform a method of vehicle anomaly location according to any embodiment of the present invention.

In a fourth aspect, the present embodiment provides a computer-readable storage medium, where the computer program is executed by the at least one processor to enable the at least one processor to execute the method for locating an abnormality of a vehicle according to any one of the embodiments of the present invention.

The embodiment of the invention provides a method, a device, equipment and a storage medium for positioning the abnormity of a vehicle, wherein the method comprises the following steps: determining target combination data corresponding to the vehicle when the vehicle is abnormal in functional operation according to at least one type of operation data generated in the process of performing functional operation on the vehicle, wherein the target combination data comprise operation problem types and characteristic data associated with the operation problem types; training a set machine learning model according to the target combination data to obtain a classification model; and positioning an abnormal operation object of the vehicle when the function operation is abnormal according to the classification model. By utilizing the technical scheme, the problem data with abnormal function operation is determined through the operation data, then the problem data is preprocessed, the model is trained based on the processed data to obtain the classification model, and the abnormal operation object influencing the decision-making effect under different problems is determined according to the classification model.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic flowchart of an anomaly locating method for a vehicle according to an embodiment of the present invention;

fig. 2 is a schematic flow chart of a vehicle abnormality positioning method according to a second embodiment of the present invention;

fig. 3 is a schematic structural diagram of an abnormality positioning device for a vehicle according to a third embodiment of the present invention;

fig. 4 is a schematic structural diagram of an electronic device according to a fourth embodiment of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "original", "target", and the like in the description and claims of the present invention and the drawings described above are used for distinguishing similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other sequences than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Example one

Fig. 1 is a flowchart of an abnormality locating method for a vehicle according to an embodiment of the present invention, where the method is applicable to locating an abnormality of the vehicle according to operation data, and the method may be executed by an abnormality locating device of the vehicle, where the abnormality locating device of the vehicle may be implemented in a form of hardware and/or software, and the device may be integrated in an electronic device.

As shown in fig. 1, the method for locating an abnormality of a vehicle according to the first embodiment may specifically include the following steps:

and S110, determining corresponding target combination data when the vehicle is abnormal in functional operation according to at least one type of operation data generated in the process of performing functional operation on the vehicle.

In this embodiment, it is necessary to perform abnormality positioning on the vehicle based on the operation data. Therefore, when the method for locating an abnormality of a vehicle according to the present embodiment is executed, at least one type of operation data generated during the operation of the vehicle is acquired. The operation data is generated in the process of performing functional operation on the vehicle, and the functional operation is required to be performed on the vehicle to obtain the operation data.

Before the vehicle is functionally operated, it is required to ensure that a designated sensor is installed on the vehicle, and the vehicle can normally operate and collect data in the driving process so as to ensure that specific problems can be found through the data subsequently. Illustratively, the sensors mainly include a camera (camera), a Global Positioning System (GPS), a laser radar (lidar), a millimeter wave radar (radar), and a Controller area network (Can).

Before the vehicle is functionally operated, each internally developed algorithm module needs to be compiled, namely, a decision-making module is calculated and deployed on a vehicle-end intelligent decision-making module, and exemplarily, the algorithm module may include a sensing module, a positioning module, a fusion module, a planning module or a control module and the like. After the deployment is ensured to be successful and can be operated, the data acquired by the sensors can be transmitted to each algorithm module in real time in a data stream form to serve as data source data. And different algorithm modules output the calculation decision results after calculation through the algorithm models of the different algorithm modules, and store the data acquired by the original sensors and the calculation decision results of the algorithm modules in the vehicle-end storage device after being associated.

And driving the vehicle to an actual running road section for running, collecting environmental data at each moment in the running process, decision making process and result data of an algorithm module, and recording abnormal conditions and problem data.

In the running process of the automatic driving algorithm, the sensor acquisition of data such as surrounding environment data, data interacted with a target object, lane line information and the like is required. In addition, in the operation process, the normal operation of the vehicle-end problem recording system needs to be ensured, and recording personnel need to record problem information (such as non-centered driving, too close distance to a target object, dragon drawing during driving and the like) and the time when problems occur in the operation process.

After the vehicle function operation is completed, the original sensor data collected by the vehicle end, the calculation result decision data of different algorithm modules and the problem data recorded in the operation process are stored in a storage device in a unified mode. And carrying out storage management on the stored data, acquiring the stored data in a server access mode and carrying out subsequent data processing calculation.

Specifically, at least one type of running data is acquired based on running data generated by functional running of the vehicle, target combination data corresponding to the abnormal functional running of the vehicle is determined according to the running data, the target combination data comprise running problem types and characteristic data associated with the running problem types, and abnormal positioning is carried out on the vehicle based on the target combination data.

In this embodiment, the operation data may include raw acquisition data acquired by the sensor device, decision data obtained by performing an operation by the algorithm decision module, and problem data corresponding to various abnormal situations. It can be understood that the functional operation of the vehicle may be abnormal in that the acquired data is abnormal due to abnormality of the acquisition devices such as the sensors, so that the decision data is abnormal, so that the functional operation of the vehicle is abnormal, or the algorithm decision module has a problem, so that the functional operation of the vehicle is abnormal due to the abnormal decision data.

The decision data refers to data calculated by an algorithm decision module according to the environmental data. For example, a non-moving obstacle is recognized 20 meters ahead according to a sensing algorithm, the sensing module recognizes the type of the object and informs the control module, and then the control module performs driving control calculation based on the recognized information, the relative position information of the obstacle and the relevant information of the lane line, for example, the operation of driving for 1 meter in a transverse direction within 5 seconds or stopping is required.

In order to determine the cause of the abnormality in the functional operation of the vehicle, the determination is made based on each operation data in the present embodiment. And (4) carrying out statistical classification on various abnormal information and abnormal phenomena, and processing each operation data. The problem occurring in the functional operation of the vehicle is statistically analyzed with data that may be associated with the problem.

First, correlation is performed on relevant data in the operational data, such as correlating raw collected data, decision data, and problem data. It can be understood that the raw collected data is automatically collected by a data collecting device installed on the vehicle, such as a sensor, so that the system can automatically record the relevant information of the raw collected data, such as the collecting time, the data type, the data value and the like of the raw collected data. The decision data are data generated by automatic calculation of an algorithm module arranged on the vehicle, so that a vehicle-end system can automatically record relevant information of the decision data, such as the generation time of the decision data. In consideration of the association of time, the original acquisition data corresponding to the same time can be associated with the decision data by taking the time stamp as the identifier. Because different sensors collect the frequency, the algorithm calculates the frequency and the problem record time all has the difference. Therefore, the data cannot be directly correlated by the time stamp, and the data can be frequency-aligned by means of frequency compression.

The problem data mainly comprises problem data which are automatically recorded and found by a system and problem data which are manually recorded in the running process of the vehicle function. For the problem data automatically recorded by the system, the system can automatically record the time and other related information, so that for the data, the timestamp can be used as an identifier, and the data obtained by associating the original collected data with the decision data is associated with the problem data recorded by the system.

For problem data recorded manually, manual recording often has hysteresis during recording, and therefore correlation can be performed based on an approximate time stamp. Illustratively, the data obtained by correlating the original acquisition data with the minimum time difference with the problem data with the decision data is correlated with the manually recorded problem.

It should be appreciated that the problem data automatically recorded by the system is normalized and does not require further processing. For the problem data recorded manually, the problem data is recorded in a text mode and has no regularity. In this embodiment, the manually recorded questions are processed to obtain classified question data. For example, the classification processing manner may be a word segmentation processing manner.

Meanwhile, since the processed operation data needs to be input into the model subsequently, discrete processing is required to be performed on the continuous operation data to obtain discrete data for inputting into the model.

For the data after the association processing and the discrete processing, each data feature input model is a singular value feature input, and only the influence of a single data feature on the final classification result can be observed when the data is finally observed. However, in the decision process of the algorithm module, the influence of the common combination of two or more characteristics on the result can exist. Therefore, in order to search for and discover such a phenomenon, it is necessary to cross features of individual independent features to construct cross features.

In this embodiment, the cross feature, the original collected data, and the decision data are used as feature data associated with each operation problem category. And taking the problem data automatically recorded by the system and the manually recorded problem data after classification as the operation problem category. And determining the operation problem categories and the characteristic data associated with the operation problem categories as target combined data.

And S120, training the set machine learning model according to the target combination data to obtain a classification model.

In this embodiment, after performing data association and preprocessing on all the operation data to obtain target combination data, modeling is performed on the target combination data through a machine learning model. Considering the case that the data has abnormal values and missing values, when selecting the model, it is necessary to select a model that is not sensitive to the missing values and the abnormal values, and it can also be understood that the model cannot be calculated due to the existence of the missing values, or the model has a large loss of stability and accuracy due to the existence of abnormal range values. Furthermore, the present embodiment is mainly used for searching and finding data features, so that it is necessary to have high interpretability of the model, and considering comprehensively, the model in the present embodiment is preferably a random forest. Specifically, a set machine learning model is trained based on the target combination data to obtain a classification model.

For example, modeling may be performed according to feature data associated with each operation problem category in the target combination data, and the feature data associated with each operation problem category is input into the model, so that the model is trained with the operation problem category as a target.

And S130, positioning an abnormal operation object when the function operation of the vehicle is abnormal according to the classification model.

The abnormal operation object may be a sensor acquisition device or an algorithm decision module. In this embodiment, according to the trained model, the contribution degree of each feature data may be determined, and the feature data related to the operation problem may be determined according to the contribution degree. And acquiring which sensor device collects the relevant characteristic data or which algorithm module is involved according to the field name and the like of the relevant characteristic data. The involved sensor devices or algorithm modules are determined as abnormal operation objects when the vehicle is in abnormal function operation. And calculating the data or algorithm decision modules which specifically influence the decision effect under different operation problem types through the model, and finally accurately positioning the problems through the data.

The embodiment of the invention provides an abnormity positioning method of a vehicle, which comprises the following steps: determining target combination data corresponding to the vehicle when the vehicle is abnormal in functional operation according to at least one type of operation data generated in the process of performing functional operation on the vehicle, wherein the target combination data comprise operation problem types and characteristic data associated with the operation problem types; training a set machine learning model according to the target combination data to obtain a classification model; and positioning an abnormal operation object of the vehicle when the function operation is abnormal according to the classification model. By means of the technical scheme, the problem data with abnormal function operation are determined through the operation data, then the problem data are preprocessed, the model is trained based on the processed data to obtain the classification model, the abnormal operation objects influencing the decision-making effect under different problems are determined according to the classification model, and compared with the prior art, abnormal operation positioning is manually performed.

Example two

Fig. 2 is a schematic flow diagram of an abnormality positioning method for a vehicle according to a second embodiment of the present invention, which is a further optimization of the second embodiment of the present invention, and in the present embodiment, a limited optimization is further performed to "determine target combination data corresponding to a vehicle when the vehicle is in an abnormal functional operation according to at least one type of operation data generated during a functional operation of the vehicle", train a set machine learning model according to the target combination data to obtain a classification model ", and perform a limited optimization to" position an operation abnormal object of the vehicle when the vehicle is in an abnormal functional operation according to the target classification model ".

As shown in fig. 2, the second embodiment provides a method for locating an abnormality of a vehicle, which specifically includes the following steps:

and S210, determining target associated data after the operation data are associated according to the operation data generated in the process of performing function operation on the vehicle.

In this step, the originally acquired data acquired by the sensor, the decision data calculated by the algorithm decision module, and the recorded problem data need to be correlated. Because the acquisition frequency of different sensors, the calculation frequency of an algorithm and the problem recording time are different. And therefore cannot be directly correlated by means of a time stamp. The frequency alignment is carried out according to the acquisition frequency of the sensor and the calculation frequency of the algorithm, and then the correlation is carried out according to the time for recording the problem and the aligned data.

Preferably, the operational data includes: raw acquisition data, decision data, and problem data. Further, the step of determining the target associated data associated with each operating data according to each operating data generated during the functional operation of the vehicle may be optimally expressed as:

a1, compressing and aligning the acquisition frequency of the original acquisition data and the calculation frequency of decision data related to the original acquisition data in a lowest frequency alignment mode to obtain the compressed original acquisition data and the decision data.

In this embodiment, the original collected data and the decision data are associated, and since the collected frequencies of different sensors and the calculation frequencies of the algorithm modules are different, the collected data frequencies of the sensors and the calculation frequencies of the algorithm modules need to be aligned. Specifically, the alignment is performed by respectively calculating each sensor and its associated algorithm module through a form of lowest frequency alignment, such as: and if the acquisition frequency of the lidar is 20HZ and the calculation frequency of the correlation algorithm module is 10HZ, reducing the acquisition frequency of the lidar to 10HZ.

Specifically, the frequency may be reduced in a data compression manner, and the compressed data may be represented as:

HZ _out ＝List(HZ _Zip )，

wherein, HZ _max Representing the maximum value of frequency HZ in the acquisition frequency and the calculation frequency _min Represents the minimum value of the acquisition frequency and the calculation frequency, and n represents a positive integer.

For example: HZ _lidar(a) ＝20HZ，HZ _lidar(b) =5HZ, then finally output at 5HZ, wherein the data HZ is output _lidar(a) And compressing to 5HZ, wherein the compression mode is that the results of the data from the 1 st frame to the 4 th frame are added, then the addition result is divided by 4 to calculate the average value, the calculated average value is used as the first output data of the data and is put into a List, then the data from the 5 th frame to the 8 th frame is calculated in the same mode, and so on, all the results are calculated, and a copy of compressed data is obtained. By using the method, all relevant data, mainly including original acquisition data and algorithm decision data, are subjected to frequency compression processing.

And b1, associating the compressed original acquisition data with the decision data by taking the timestamp as an identifier to obtain first associated data.

Specifically, the compressed original collected data and the decision data are correlated, the compressed data with the same timestamp are correlated, and the correlated data are recorded as first correlated data.

And c1, determining target related data according to the first related data and the problem data.

In this step, the compressed original collected data and the first associated data determined by the decision data are associated with the problem data, and the associated data are determined as target associated data. Here, the problem data includes abnormal data automatically registered by the system and manually recorded problem data. Similarly, the first association data and the question data may be associated by a time stamp.

Further, determining target associated data according to the first associated data and the problem data, including:

and c11, classifying the problem data to obtain the operation problem category.

Specifically, the recording mode of the problem data is obtained, and the recording mode comprises system recording and manual recording. The mode of the system problem data can be understood as that a judgment condition is preset in the system, the judgment is triggered when the triggering condition is met, and the found problem is automatically recorded through the system. The manual question data is obtained by manually recording the found questions during the running process of the vehicle, and exemplary manual recording questions are as follows: "the high-speed vehicle passes through the left front of this car, and this problem frequently takes place, influences normal operating in this car unusual snub brake, hard braking.

And if the recording mode is manual recording, classifying the problem data to obtain the operation problem category. In this embodiment, if the recording mode of the problem found in the operation process is manual recording, the recording may be irregular and the time lag may occur. Therefore, after the problem data is collected, all the problem data needs to be uniformly classified. The Processing and classifying method here performs word segmentation on all the recorded problems by using a word segmentation matching method in natural Language, and the word segmentation method here mainly uses a public tool method, such as Jieba word segmentation, han Language Processing package (HanLP), and the like, where the word segmentation mode includes a precise mode, a full mode, a search engine mode, and the like. Preferably, a precision mode is used, for example: "the high-speed vehicle passes through the left place ahead of this car, and this problem frequently takes place, influences normal operating in this car unusual snub brake, sudden braking", this sentence is the specific recording problem, obtains following result through the participle: "high speed/vehicle/passing/vehicle/front left/,/vehicle/abnormal/snub/,/hard brake/,/problem/frequent/occurrence/,/impact/normal/running". Matching to: "/snub/,/hard brake/" keyword. Finally, this type of word is classified as an "abnormal braking" type. By the method, all the manually recorded problem data are classified to obtain the operation problem category.

And if the recording mode is system recording, the problem data is used as the operation problem type. It is clear that the problem data recorded by the system may include information related to the problem data, such as the time when the problem occurred and the information record when the abnormality was found, which are relatively normalized records, and thus do not need to be processed excessively. The abnormal data automatically recorded by the vehicle system is classified through an automatic implementation mode, and therefore further processing is not needed.

And c12, determining target associated data according to the minimum time difference between the timestamp of the operation problem category and the timestamp of the first associated data.

Specifically, the first association data is associated with the running problem category through a timestamp. Since the data in the first associated data is subjected to compression processing, it may occur that a part of the data cannot be associated by the time stamp. Another way of data association will be used here to approximate the association of the timestamps. The method comprises the following specific steps: and associating the recording time of the problem data corresponding to the operation problem type with the time of the compressed first associated data, and if the data can be associated, directly associating. If the correlation can not be carried out, the time difference between the time stamp of the problem data and each time stamp of the compressed data is calculated. And then selecting the minimum value of the calculated time difference as the occurrence time of the associated problem data, marking the data, adding a new row of data to map and match the marked data with the operation problem category, and determining the matched data as target associated data.

And S220, obtaining numerical target associated data by combining a preset text numerical mapping relation according to the target associated data.

If the classified target data in the target associated data is text type data, text-to-numerical mapping needs to be performed on the text type data, and numerical conversion needs to be performed on all text information. Exemplary, classification target data: recognizing lane line colors including white, yellow, blue, orange, etc., since the model cannot directly use text type, white is mapped to number 1, yellow to number 2, blue to number 3, orange to number 4, etc. And performing numerical mapping on all the other phase text type data. In this embodiment, the non-classified target data used in the present embodiment are all numerical type data when the data is stored and recorded, so that further conversion is not required. And finally, converting all the target associated data into numerical target associated data.

And S230, obtaining cross data after feature crossing according to the numerical target associated data and combining a preset feature construction mode, and assigning values to the cross data.

And when the data is finally observed, only the influence of the single characteristic data on the final classification result can be observed. However, in the algorithm module decision making process, the result is influenced by the condition that two or more characteristics are combined together. Therefore, in order to search for and discover such a phenomenon, it is necessary to cross features of the individual feature data to construct cross features.

Specifically, feature cross processing is performed on the numerical target associated data in a preset feature construction mode to obtain cross data, and the cross data is assigned.

Further, the step of obtaining the crossed data and assigning values to the crossed data according to the numerical target associated data and by combining a preset feature construction mode is optimally expressed as follows:

and a2, extracting a set number of related feature data from the numerical target related data, and performing discrete processing on the feature data to obtain a discrete feature data set.

In consideration of the fact that the number of basic feature data in the numerical target associated data is large, feature cross calculation complexity and calculation efficiency are considered when feature construction is carried out, feature cross dimensionality should not exceed three-dimensional feature cross combination, otherwise, too large calculation amount can result in incapability of calculation or too long time consumption, and after too many features are generated, overfitting is caused due to too many features during model calculation. Preferably, the set number is 3.

Because of the existence of a large amount of continuous feature data, if the feature data is a continuous value, the number of the feature data is very large, which is not beneficial to model training, and the feature data needs to be subjected to discrete processing, such as barrel separation processing. In this step, the bucket dividing method is mainly based on the data range to divide the data into different types, for example, the vehicle speed range is 0-130, at this time, the vehicle speed is divided into 6 buckets, 0-20, 20-40, 40-60, 60-80, 80-100, and more than 100, and then the data in different value ranges are classified into the bucket and attached with the bucket dividing identifier. Such as 0-20 for 1, 20-40 for 2, and so on, to facilitate the use of feature intersections to construct features.

And b2, carrying out characteristic construction on the characteristic data set in a characteristic phase-AND mode to obtain cross data.

In this step, the main characteristic structural modes are the structural modes of the characteristic phase pair, for example: it is recognized that there are non-motor vehicles around and at the same time, it is recognized that the intersection with the traffic light is on, and the traffic light is green, at which time the new intersection data is that there are non-motor vehicles around and at the intersection and is green.

And c2, assigning the cross data meeting the preset parallel condition as one, and assigning the cross data not meeting the preset parallel condition as zero.

Exemplarily, the new intersection at this time is characterized by the presence of non-motor vehicles around and at the intersection and is green =1, and the remaining features not satisfying the above-described parallel condition are set to 0; by analogy, all relevant feature data are calculated.

S240, taking the numerical target related data and the cross data as feature data, and taking the determined operation problem types and the feature data related to the operation problem types as corresponding target combination data when the vehicle is in abnormal function operation.

Specifically, the numerical target associated data and the cross data are both used as feature data, and it is clear that the target associated data includes associated original collected data, decision data, and the like. And the determined operation problem types and the characteristic data associated with the operation problem types are used as corresponding target combination data when the vehicle is in abnormal function operation.

And S250, carrying out model construction on the target combined data through a random forest model.

Specifically, based on the correlated data, the model can be constructed by constructing a classification prediction model. According to business needs, more operation problem categories exist in the operation process, so that the model prediction requirements can be met only by constructing a multi-classification model. Data preprocessing is required before the model is built.

The preprocessing mainly comprises data cleaning, and because the phenomenon in the actual data acquisition and calculation decision process is unknown, how the data can be represented, and if the data is filled in by follow-up manual filling, the representation of the original data is influenced due to improper filling mode. The missing data in each field cannot be padded and repaired in a data padding mode, and only can be deleted or processed through machine learning automatic processing capacity. In the embodiment, the emphasis is on finding defects of the algorithm decision software, so that missing data is not filled or deleted, abnormal data in the missing data only needs to be judged and does not need to be processed, and otherwise problem data cannot be correctly reflected through the data.

The selection of the classification model needs to comprehensively consider the influence of a missing value, the influence of an abnormal value, model interpretability and the like. An exemplary, neighbor node algorithm (K-Neares tNeighbor, KNN) is characterized by: the influence of the missing value and the influence of the abnormal value, and the interpretability of the model, and a Decision Tree (Decision Tree) has the characteristics of influence of the missing value and the abnormal value, and interpretability of the model. The random forest is characterized in that the influence of a missing value is small, the influence of an abnormal value is small, and a model can be explained. In this embodiment, a random forest is preferably used as the classification model. Specifically, the correlated data is modeled by using a random forest model.

In this embodiment, considering that an abnormal value and a missing value may exist in the target combined data, a model that is not sensitive to the missing value and the abnormal value needs to be selected when selecting the model, that is, a situation that the model cannot be calculated due to the existence of the missing value or the stability and accuracy of the model are greatly lost due to the existence of an abnormal range value is avoided. Furthermore, the present embodiment is mainly used for the exploration and discovery of feature data, so that it is required to have high interpretability of the model, and the model is mainly selected as a random forest in comprehensive consideration.

And S260, training the model by taking the feature data in the target combined data as input and the operation problem category associated with the feature data as a target to obtain a classification model.

Specifically, after the associated data is modeled by using a random forest model, feature data in the target combination data is input into the model, and the operation problem category in the target combination data is used as a model training target.

And S270, determining the contribution degree of each feature data to the classification model according to the classification model.

Specifically, according to the classification model, the contribution degree of all the features of the model training to the model is respectively calculated.

And S280, positioning an abnormal operation object when the vehicle is abnormal in function operation according to the contribution degree.

Specifically, the abnormal operation object of the vehicle when the function operation is abnormal is positioned according to the contribution degree of each feature data to the classification model.

Further, according to the contribution degree, positioning an abnormal operation object of the vehicle when the function operation is abnormal comprises the following steps:

and a3, determining the feature data of which the total contribution degree is larger than a set threshold before the contribution degree is decreased from large to small as the target feature data.

Specifically, after the contribution degrees of the feature data are obtained, the feature data are sorted from large to small according to the contribution degrees. For the convenience of observation, the features in which the calculated contribution degree is 0 are removed.

Preferably, the set threshold may take a value of 95%. Specifically, by means of the cumulative calculation of the contribution degrees, the feature data of which the sum of the corresponding TopN contribution degrees exceeds 95% is taken as the target feature data.

And b3, determining data acquisition equipment corresponding to the field name of the target characteristic data and an algorithm decision module.

Specifically, a field name and a corresponding value of the target feature data are determined, and data acquisition equipment and corresponding calculation decision software corresponding to the field name are determined.

And c3, taking the data acquisition equipment and the calculation decision software as an abnormal operation object when the vehicle is in abnormal function operation.

Specifically, the determined data acquisition equipment and the determined calculation decision software are used as an abnormal operation object when the vehicle is abnormal in function operation. Illustratively, the corresponding sensor used for acquiring the data and an algorithm decision module used for calculating the data, such as a control module, are determined according to the name of the field of the target characteristic data.

Further, after determining the data feature with the sum of the previous continuous number contribution degrees greater than the set threshold as the target feature data, the method further includes: constructing an interpreter of a classification model based on a set algorithm; and generating an interpretation information table and an effect graph through the interpreter so as to explain the contribution degree of each target data characteristic to each classification problem.

Considering that some of the located features may be cross-features, it is necessary to disassemble the combined features to facilitate locating the most fundamental problem of the data, and some problems may be caused by that the features are not obvious before the combination of the individual features, but the features are obvious after the cross-combination of the features. Because the features obtained by the random forest calculation are importance degrees, the influence of the features on the result is not known to be positive or negative, and the result is relatively important for finding problem positioning. At this point the feature interpretation tool sharp needs to be used.

A random forest classification interpreter is constructed through Shap, the TopN characteristic which is screened through characteristics is mainly adopted at the moment, and the problem is taken as a target. After training, generating a model interpretation information table and a corresponding effect graph through a shape model interpretation function. And interpreting the influence degree of each characteristic on the target according to the positive and negative influences of the effect diagram and the influence degrees of different ranges. Wherein, if the calculated characteristic is negative, it means that the smaller the characteristic value, the more likely the problem or abnormality occurs. If the calculated characteristic is positive, it means that the larger the characteristic is, the more likely a problem or abnormality occurs.

According to the embodiment, the data acquired by various sensors in the operation process are analyzed, and the abnormal sensor information and the abnormal data information of the sensor acquisition part can be accurately identified based on the data detection and judgment capacity. And detecting and identifying abnormal data according to decision results generated by calculation of each algorithm decision module in automatic driving operation and effective ranges of data generated by different modules. The abnormal information triggered in the operation verification process in the automatic driving operation and the abnormal phenomenon discovered by the operator can be recorded. And various abnormal information and abnormal phenomena are counted and classified, the data are modeled through a machine learning model, the data or algorithm decision modules which specifically influence the decision effect under different problems are calculated through the model, and finally the problems are accurately positioned through the data.

EXAMPLE III

Fig. 3 is a schematic structural diagram of an abnormality locating device for a vehicle according to a third embodiment of the present invention, which is applicable to locating an abnormality of a vehicle according to operation data, and the device may be implemented in the form of hardware and/or software and is generally integrated in an electronic device. As shown in fig. 3, the apparatus includes: data determination module 31, model determination module 32, anomaly localization module 33, wherein,

the data determining module 31 is configured to determine, according to at least one type of operation data generated in a process of performing functional operation on a vehicle, target combination data corresponding to the vehicle when the functional operation is abnormal, where the target combination data includes operation problem types and feature data associated with the operation problem types;

the model determining module 32 is configured to train a set machine learning model according to the target combination data to obtain a classification model;

and an anomaly positioning module 33, configured to position an abnormal operation object of the vehicle when the function operation is abnormal according to the classification model.

Optionally, the data determining module 32 may include:

the associated data determining unit is used for determining target associated data after the association of the operating data according to the operating data generated in the process of performing function operation on the vehicle;

the numerical data determining unit is used for obtaining numerical target associated data by combining a preset text numerical mapping relation according to the target associated data;

the intersection acquisition unit is used for acquiring intersection data after feature intersection and assigning values to the intersection data according to the numerical target associated data by combining a preset feature construction mode;

and the combination determining unit is used for taking the numerical target related data and the cross data as characteristic data, and taking the determined operation problem type and the characteristic data related to each operation problem type as corresponding target combination data when the vehicle is in abnormal function operation.

Optionally, the operational data comprises: original collected data, decision data and problem data;

the associated data determining unit may include:

the first acquisition subunit is used for compressing and aligning the acquisition frequency of the original acquisition data and the calculation frequency of the decision data related to the original acquisition data in a lowest frequency alignment mode to acquire the compressed original acquisition data and the decision data;

the second obtaining subunit is used for associating the compressed original acquisition data with the decision data by taking the timestamp as an identifier to obtain first associated data;

and the association determining subunit is used for determining the target associated data according to the first associated data and the problem data.

Optionally, the association determining subunit is specifically configured to:

classifying the problem data to obtain the category of the operation problem;

and determining the target associated data according to the minimum time difference between the timestamp of the operation problem category and the timestamp of the first associated data.

Optionally, the intersection obtaining unit is specifically configured to:

extracting a set number of related feature data from the numerical target associated data, and performing discrete processing on the feature data to obtain a discrete feature data set;

carrying out feature construction on the feature data set in a feature phase-and-phase mode to obtain cross data;

and assigning the cross data meeting the preset parallel condition as one, and assigning the cross data not meeting the preset parallel condition as zero.

Optionally, the model determining module 32 may be specifically configured to:

the model construction unit is used for carrying out model construction on the target combined data through a random forest model;

and the classification model determining unit is used for training the model by taking the characteristic data in the target combined data as input and the operation problem category associated with the characteristic data as a target to obtain a classification model.

Optionally, the anomaly locating module 33 may include:

the contribution degree determining unit is used for determining the contribution degree of each feature data to the classification model according to the classification model;

and the abnormity positioning unit is used for positioning an abnormal operation object of the vehicle when the function operation is abnormal according to the contribution degree.

Further, the anomaly locating unit is specifically configured to:

determining feature data with the sum of the contribution degrees of the continuous numbers larger than a set threshold before the contribution degrees are reduced from large to small as target feature data;

determining data acquisition equipment and an algorithm decision module corresponding to the field name of the target characteristic data;

and taking the data acquisition equipment and the algorithm decision module as an abnormal operation object when the vehicle is in abnormal function operation.

Optionally, the apparatus further includes an interpretation module, specifically configured to:

constructing an interpreter of a classification model based on a set algorithm;

and generating an interpretation information table and an effect graph through the interpreter so as to be used for interpreting the contribution degree of each target data characteristic to each classification problem.

The vehicle abnormity positioning device provided by the embodiment of the invention can execute the vehicle abnormity positioning method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.

Example four

Fig. 4 is a schematic structural diagram of an electronic device according to a fourth embodiment of the present invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.

As shown in fig. 4, the electronic device 40 includes at least one processor 41, and a memory communicatively connected to the at least one processor 41, such as a Read Only Memory (ROM) 42, a Random Access Memory (RAM) 43, and the like, wherein the memory stores a computer program executable by the at least one processor, and the processor 41 may perform various suitable actions and processes according to the computer program stored in the Read Only Memory (ROM) 42 or the computer program loaded from a storage unit 48 into the Random Access Memory (RAM) 43. In the RAM 43, various programs and data necessary for the operation of the electronic apparatus 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other via a bus 44. An input/output (I/O) interface 45 is also connected to the bus 44.

A plurality of components in the electronic device 40 are connected to the I/O interface 45, including: an input unit 46 such as a keyboard, a mouse, etc.; an output unit 47 such as various types of displays, speakers, and the like; a storage unit 48 such as a magnetic disk, optical disk, or the like; and a communication unit 49 such as a network card, modem, wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

Processor 41 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 41 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, and so forth. Processor 41 performs the various methods and processes described above, such as a vehicle anomaly locating method.

In some embodiments, the method of vehicle anomaly location may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 48. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 40 via the ROM 42 and/or the communication unit 49. When the computer program is loaded into RAM 43 and executed by processor 41, one or more steps of the above-described method of vehicle anomaly location may be performed. Alternatively, in other embodiments, processor 41 may be configured to perform the vehicle anomaly locating method by any other suitable means (e.g., by way of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

A computer program for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be performed. A computer program can execute entirely on a machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of the present invention, a computer-readable storage medium may 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. A computer readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer readable storage medium may be a machine readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here 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 a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the electronic device. Other kinds of devices may also be used to provide for interaction with a 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, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end 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 Networks (LANs), wide Area Networks (WANs), blockchain networks, and the internet.

The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical host and VPS service are overcome.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present invention may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired result of the technical solution of the present invention can be achieved.

The above-described embodiments should not be construed as limiting the scope of the invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for locating an abnormality of a vehicle, comprising:

2. The method according to claim 1, wherein the determining the target combination data corresponding to the vehicle with abnormal functional operation according to at least one type of operation data generated in the process of performing functional operation on the vehicle comprises:

determining target associated data after association of the operation data according to the operation data generated in the process of performing functional operation on the vehicle;

according to the target associated data, combining a preset text numerical value mapping relation to obtain numerical target associated data;

according to the numerical target associated data, combining a preset characteristic construction mode to obtain cross data after characteristic cross and assigning values to the cross data;

and taking the numerical target associated data and the cross data as feature data, and taking the determined operation problem types and the feature data associated with each operation problem type as corresponding target combined data when the vehicle is in abnormal function operation.

3. The method of claim 2, wherein the operational data comprises: original collected data, decision data and problem data;

the step of determining target associated data associated with each operation data according to each operation data generated in the process of performing functional operation on the vehicle comprises the following steps:

compressing and aligning the acquisition frequency of the original acquisition data and the calculation frequency of decision data related to the original acquisition data in a lowest frequency alignment mode to obtain compressed original acquisition data and decision data;

correlating the compressed original collected data with the decision data by taking a timestamp as an identifier to obtain first correlated data;

and determining target associated data according to the first associated data and the problem data.

4. The method of claim 3, wherein determining target correlation data based on the first correlation data and the problem data comprises:

classifying the problem data to obtain an operation problem category;

and determining target associated data according to the minimum time difference between the timestamp of the operation problem category and the timestamp of the first associated data.

5. The method according to claim 2, wherein the obtaining cross data after feature crossing and assigning values to the cross data according to the numerical type target associated data in combination with a preset feature construction manner comprises:

carrying out feature construction on the feature data set in a feature AND mode to obtain cross data;

6. The method according to claim 1, wherein the training of the set machine learning model according to the target combination data to obtain a classification model comprises:

carrying out model construction on the target combination data through a random forest model;

and training the model by taking the characteristic data in the target combination data as input and taking the operation problem category associated with the characteristic data as a target to obtain the classification model.

7. The method of claim 1, wherein said locating an abnormal operation object of the vehicle at the time of the abnormal operation of the function according to the classification model comprises:

determining the contribution degree of each feature data to the classification model according to the classification model;

and positioning an abnormal operation object of the vehicle when the function operation is abnormal according to the contribution degree.

8. The method according to claim 7, wherein the locating the abnormal operation object of the vehicle when the function operation is abnormal according to the contribution degree comprises:

determining feature data with the sum of the contribution degrees of the continuous numbers larger than a set threshold value before the contribution degrees are decreased from large as target feature data;

9. The method according to claim 8, wherein after determining, as the target feature data, the data feature whose sum of the contribution degrees of the consecutive numbers from large to small is greater than the set threshold, the method further comprises:

constructing an interpreter of the classification model based on a set algorithm;

10. An abnormality positioning device for a vehicle, characterized by comprising:

11. An electronic device, characterized in that the electronic device comprises:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the first and the second end of the pipe are connected with each other,

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the method of vehicle anomaly location of any one of claims 1-9.

12. A computer-readable storage medium storing computer instructions for causing a processor to implement the method for locating an abnormality of a vehicle according to any one of claims 1 to 9 when executed.