Automatic driving data mining method and device

Abstract

The application provides an automatic driving data mining method and device, and the method comprises the following steps: acquiring automatic driving data; inputting the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; judging whether the vehicle state is changed according to the vehicle predicted state and a previously acquired current vehicle state to obtain a first vehicle state switching result; comparing the first vehicle state switching result with a second vehicle state switching result acquired offline based on the automatic driving data to obtain a first comparison result, determining whether the first comparison result meets a first preset condition, and determining to return the automatic driving data based on the fact that the first comparison result does not meet the first preset condition. The application realizes the migration of offline learning ability to online learning ability, improves the learning ability of online learning, and makes the online learning system and the offline learning system evolve cooperatively, so that the two systems are more intelligent.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, in particular to an automatic driving data mining method and device. BACKGROUND

[0002] With the popularization of L2 and above automatic driving technology, data driving has become an industry consensus, however, the massive data has brought great challenges to data storage, transmission and analysis, and how to improve the quality of the data on disk, automatically process the massive data generated every day, and reduce human intervention as much as possible, is still a pain point in the industry.

[0003] At present, when mining automatic driving data, a plurality of source sensors are arranged on a vehicle to collect massive data in the driving process of the vehicle, and then relevant personnel mine the corresponding required scene data from the massive data collected above to comprehensively test the automatic driving system. However, the above mining method needs manual collection and mining, which is time-consuming and laborious and has poor intelligence. SUMMARY

[0004] The present application provides an automatic driving data mining method and device to solve the defect of manual mining of massive automatic driving data in the prior art, thereby improving the data mining efficiency.

[0005] The present application provides an automatic driving data mining method, comprising: obtaining automatic driving data; inputting the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training the vehicle training data and the vehicle state true value corresponding to the vehicle training data; determining whether the vehicle state changes according to the vehicle predicted state and the previously obtained current vehicle state to obtain a first vehicle state switching result; comparing the first vehicle state switching result with a second vehicle state switching result obtained offline based on the automatic driving data in advance to obtain a first comparison result, determining whether the first comparison result meets a first preset condition, and based on the non-compliance, determining to return the automatic driving data.

[0006] According to the automatic driving data mining method provided by the present application, after obtaining the vehicle predicted state output by the vehicle state detection model, the method comprises: obtaining a first vehicle driving mode switching decision according to the vehicle predicted state and the previously obtained current driving mode; comparing the first vehicle driving mode switching decision with a second vehicle driving mode switching decision obtained offline based on the automatic driving data in advance to obtain a second comparison result, determining whether the second comparison result meets a second preset condition, and based on the non-compliance, determining to return the automatic driving data.

[0007] The automatic driving data mining method provided by the present application comprises: determining to switch a current driving mode according to the first vehicle driving mode switching decision; determining to return the automatic driving data based on whether a driver takes over when the current driving mode is switched; and determining to return the automatic driving data if it is determined that the driving mode after switching does not meet the expectation after completing the driving mode switching based on the driver not taking over when the current driving mode is switched.

[0008] The automatic driving data mining method provided by the present application comprises: determining to switch a current driving mode according to the first vehicle driving mode switching decision; determining to return the automatic driving data based on whether a driver takes over when the current driving mode is switched; and determining to return the automatic driving data if it is determined that the driving mode after switching does not meet the expectation after completing the driving mode switching based on the driver not taking over when the current driving mode is switched.

[0009] The automatic driving data mining method provided by the present application comprises: determining to switch a current driving mode according to the first vehicle driving mode switching decision; determining to return the automatic driving data based on whether a driver takes over when the current driving mode is switched; and determining to return the automatic driving data if it is determined that the driving mode after switching does not meet the expectation after completing the driving mode switching based on the driver not taking over when the current driving mode is switched.

[0010] The automatic driving data mining method provided by the present application comprises: determining to switch a current driving mode according to the first vehicle driving mode switching decision; determining to return the automatic driving data based on whether a driver takes over when the current driving mode is switched; and determining to return the automatic driving data if it is determined that the driving mode after switching does not meet the expectation after completing the driving mode switching based on the driver not taking over when the current driving mode is switched.

[0011] The automatic driving data mining method provided by the present application comprises: determining to switch a current driving mode according to the first vehicle driving mode switching decision; determining to return the automatic driving data based on whether a driver takes over when the current driving mode is switched; and determining to return the automatic driving data if it is determined that the driving mode after switching does not meet the expectation after completing the driving mode switching based on the driver not taking over when the current driving mode is switched.

[0012] The application further provides an automatic driving data mining device, comprising: a data acquisition module configured to acquire automatic driving data;

[0013] a state detection module configured to input the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training the vehicle training data and a vehicle state true value corresponding to the vehicle training data; a state judgment module configured to judge whether a vehicle state changes according to the vehicle predicted state and a previously acquired current vehicle state to obtain a first vehicle state switching result; and a data return module configured to compare the first vehicle state switching result with a second vehicle state switching result acquired offline based on the automatic driving data to obtain a first comparison result, determine whether the first comparison result meets a first preset condition, and determine to return the automatic driving data based on a failure to meet the first preset condition.

[0014] The application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the automatic driving data mining method according to any one of the above when executing the program.

[0015] The application further provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executable on a processor to implement the steps of the automatic driving data mining method according to any one of the above.

[0016] The application further provides a computer program product comprising a computer program, wherein the computer program is executable on a processor to implement the steps of the automatic driving data mining method according to any one of the above.

[0017] The automatic driving data mining method and device provided by the application determine a first vehicle state judgment result by detecting the vehicle state of the vehicle state detection model, combine the first vehicle state judgment result acquired online with a first preset weight to obtain a first state, combine a second vehicle judgment result acquired offline in advance with a second preset weight to obtain a second state, and guide the acquisition of the first state online according to the second state acquired offline, thereby realizing the migration of the offline learning ability to the online learning ability, improving the learning ability of online learning, realizing the collaborative evolution of the online learning system and the offline learning system, and making both more intelligent. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0019] Figure 1 is one of the flowcharts of the automatic driving data mining method provided by the present application;

[0020] Figure 2 is another flowchart of the automatic driving data mining method provided by the present application;

[0021] Figure 3 is a structural schematic diagram of the automatic driving data mining device provided by the present application;

[0022] Figure 4 is a structural schematic diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0023] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0024] Figure 1 A flowchart of an automatic driving data mining method is shown. The method comprises:

[0025] S11, acquiring automatic driving data;

[0026] S12, inputting the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training vehicle training data and vehicle state true value corresponding to the vehicle training data;

[0027] S13, judging whether the vehicle state changes according to the vehicle predicted state and the previously acquired current vehicle state to obtain a first vehicle state switching result;

[0028] S14, comparing the first vehicle state switching result with a second vehicle state switching result acquired offline based on the automatic driving data to obtain a first comparison result, determining whether the first comparison result meets a first preset condition, and determining to return the automatic driving data based on the non-compliance.

[0029] It should be noted that S1N in the specification does not represent the order of the automatic driving data mining method, and the following will be specifically combined Figure 2 The automatic driving data mining method of the present application is described.

[0030] Step S11, obtaining automatic driving data. In this embodiment, the automatic driving data is obtained, including: obtaining driving images and vehicle state data, the vehicle state data including vehicle steering and acceleration and deceleration information, etc.; and / or, obtaining radar point cloud data and vehicle state data, the vehicle state data including vehicle steering and acceleration and deceleration information, etc.; and / or, obtaining integrated navigation data and vehicle state data, the vehicle state data including vehicle steering and acceleration and deceleration information. Specifically, the driving images are obtained, including: based on the camera, obtaining the driving images in the target time period; or, obtaining the video stream of the target area; based on the video stream, extracting the driving images in the target time period.

[0032] It should be added that based on the video stream, the driving images in the target time period are extracted, including: based on the video stream, collecting a certain number of driving images before and after the current frame driving image, for example, collecting ten frames of driving images before and after the current frame driving image, so as to avoid the case that the judgment result and the detection result based on the single frame driving image are poor, and the specific number of driving images obtained can be set according to the actual detection needs, which is not limited further here.

[0033] In addition, the driving image is an image taken based on the driving area in the vehicle. The vehicle can be a vehicle, a ship or an airplane, etc. for carrying people or carrying goods, wherein the vehicle can be a private car or an operating vehicle, such as a shared car, a network car, a taxi, a bus, a school bus, a truck, a passenger car, a train, a subway and a tram, etc.

[0034] Step S12, inputting the automatic driving data into the vehicle state detection model to obtain the vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training the vehicle training data and the vehicle state true value corresponding to the vehicle training data. The vehicle predicted state includes lane line prediction result, dynamic obstacle prediction result, lane boundary prediction result, positioning prediction result, route planning result, throttle prediction result and brake prediction result.

[0035] In this embodiment, the vehicle state detection model is trained, including: obtaining the vehicle training data and the vehicle state true value corresponding to the vehicle training data; taking the vehicle training data as the input data for training, taking the vehicle state true value as the label for training, training the to-be-trained model to obtain the vehicle state detection model for detecting the vehicle state.

[0036] It should be noted that the to-be-trained model can be an existing network built in the training model, which usually includes a network structure, or other networks specified by a user, such as an XGBOOST algorithm. In addition, the vehicle state true value can be determined according to the vehicle state actually needed to be detected, such as a lane line, a dynamic obstacle, a lane boundary, a positioning result, route planning, an accelerator, and a brake.

[0037] In step S13, whether the vehicle state changes is determined according to the vehicle predicted state and the previously acquired current vehicle state, and a first vehicle state switching result is obtained.

[0038] In the embodiment, obtaining the vehicle state determination result includes: determining whether the vehicle driving state is stable according to the vehicle predicted state and the previously acquired current vehicle state; if the vehicle does not appear a dragon or an unexpected steering situation, it is determined that the vehicle driving state is stable; if the vehicle appears a dragon or an unexpected steering situation, it is determined that the vehicle driving state is unstable. It should be noted that the dragon refers to the automatic driving vehicle swinging back and forth on the road.

[0039] In step S14, a first comparison result is obtained by comparing the first vehicle state switching result and a second vehicle state switching result acquired based on the automatic driving data offline in advance, it is determined whether the first comparison result meets a first preset condition, and based on not meeting the first preset condition, it is determined to return the automatic driving data.

[0040] It should be noted that the first vehicle state switching result is assigned a first preset weight, and the second vehicle state switching result is assigned a second preset weight, so that the first vehicle state switching result and the second vehicle state switching result after being assigned weights are compared to determine whether to automatically return the automatic driving data.

[0041] As can be seen, the first preset weight and the second preset weight can be determined according to the learning ability of the actual vehicle state detection model. When the learning ability of the vehicle state detection model is poor in the early stage, the first preset weight needs to be less than the second preset weight, so that the determination of data return mainly depends on the judgment result of prior experience or historical driving data. With the enhancement of the learning ability of the vehicle state detection model, the first preset weight can be appropriately increased and the second preset weight can be appropriately reduced. Specifically, the first preset weight and the second preset weight can be determined according to the actual design scene and the actual acquired automatic driving data, which is not limited further herein.

[0042] Further, the first preset condition can be whether a difference between the first vehicle state switching result with the first preset weight and the second vehicle state switching result with the second preset weight is within a first preset range, and based on not being within the first preset range, determining that the first preset condition is not met. In actual design, the first preset range can be set according to the actual first vehicle state switching result, the second vehicle state switching result, the first preset weight and the second preset weight, which is not further defined here.

[0043] For example, when the learning data of the vehicle detection model is less than 500h, the first preset weight is less than the second preset weight, when the learning data of the vehicle detection model is within 500h-1000h, the first preset weight and the second preset weight can be equal, and when the learning data of the vehicle detection model is greater than 1000h, the first preset weight is greater than the second preset weight.

[0044] In an optional embodiment, before determining whether the first preset condition is met according to the first vehicle state switching result and the first preset weight, and the second vehicle state switching result and the second preset weight obtained offline based on the automatic driving data in advance, it comprises: judging whether the vehicle driving state is stable according to the automatic driving data combined with prior experience or historical driving data judgment result, to obtain a second vehicle state judgment result. It should be noted that in actual processing, the judgment can also be based on the experience of the driver, which can be set according to actual design requirements, and is not further limited here.

[0045] In an optional embodiment, after obtaining the vehicle predicted state output by the vehicle state detection model, it comprises: obtaining a corresponding first vehicle driving mode switching decision according to the vehicle predicted state and the previously obtained current driving mode; comparing the first vehicle driving mode switching decision with the second vehicle driving mode switching decision obtained offline based on the automatic driving data in advance to obtain a second comparison result, determining whether the second comparison result meets a second preset condition, and based on not meeting, determining to return the automatic driving data.

[0046] Further, the first vehicle driving mode switching decision is given a third preset weight, and the second vehicle driving mode switching decision is given a fourth preset weight, so that the first vehicle driving mode switching decision and the second vehicle driving mode switching decision with the respective weights are compared to determine whether to automatically return the automatic driving data.

[0047] It should be noted that the third preset weight can refer to the first preset weight setting, and the fourth preset weight can refer to the second preset weight setting, which will not be repeated here. In addition, whether the difference between the first vehicle driving mode switching decision given the third preset weight and the second vehicle driving mode switching decision given the fourth preset weight is within the second preset range can be determined based on whether the second preset condition is not met. In the actual design process, the second preset range can be set according to the actual first vehicle driving mode switching decision, the second vehicle driving mode switching decision, the third preset weight and the fourth preset weight, which will not be further set here.

[0048] In addition, before determining whether to switch the driving mode according to the vehicle predicted state and the previously obtained current driving mode, it includes: determining whether to switch the current driving mode according to the automatic driving data combined with prior experience or historical driving data, to obtain a second vehicle driving mode switching decision.

[0049] Further, after obtaining the corresponding first vehicle driving mode switching decision, it includes: determining to switch the current driving mode according to the first vehicle driving mode switching decision; determining to return the automatic driving data based on the driver taking over when switching the current driving mode; based on the driver not taking over when switching the current driving mode, after completing the driving mode switching, if it is determined that the switched driving mode does not meet the expectation, it is determined to return the automatic driving data.

[0050] Further, obtaining the corresponding first vehicle driving mode switching decision includes: if it is determined according to the vehicle predicted state that the vehicle driving state is stable, then: in the case that the current vehicle driving mode is the lane center keeping LCC mode, if the vehicle control behavior determined based on the high-precision map is consistent with the driving state in the current LCC mode (no unexpected steering or drawing a dragon will occur when switching to the Full AD mode), the current vehicle driving mode is switched to the automatic driving Full AD mode; in the case that the current vehicle driving mode is the Full AD mode, if the vehicle control behavior determined based on the high-precision map is inconsistent with the driving state in the current Full AD mode, the current vehicle driving mode is switched to the LCC mode.

[0051] In addition, obtaining the corresponding first vehicle driving mode switching decision also includes: if it is determined according to the vehicle predicted state that the vehicle driving state is unstable, the current vehicle driving mode is switched to the fallback mode for exiting the automatic driving mode.

[0052] In an optional embodiment, after obtaining the vehicle predicted state output by the vehicle state detection model, the method further comprises: determining whether to return data according to the vehicle predicted state, the preset data return scene and the preset data return rule, to obtain a first data return decision; determining whether to meet a third preset condition according to the first data return decision and a fifth preset weight, and a second data return decision based on the automatic driving data obtained offline in advance and a sixth preset weight, and determining to return the automatic driving data based on not meeting the third preset condition.

[0053] It should be noted that the fifth preset weight can be set with reference to the first preset weight, and the sixth preset weight can be set with reference to the second preset weight, which will not be repeated here. In addition, the difference between the first data return decision assigned with the fifth preset weight and the second data return decision assigned with the sixth preset weight can be determined whether to meet a third preset range, and the third preset condition is determined not to be met based on not meeting the third preset range. In the actual design process, the third preset range can be set according to the actual first data return decision, the second data return decision, the fifth preset weight and the sixth preset weight, which will not be further set here.

[0054] It should be noted that in the case where the first return decision and the second return decision do not meet the third preset condition, as long as one of the decisions requires data return, data return should be performed.

[0055] In an optional embodiment, before determining whether to return data according to the vehicle predicted state, the preset data return scene and the preset data return rule, the method further comprises: determining whether to return data according to the automatic driving data, combined with prior experience or historical driving data, to obtain a second data return decision.

[0056] In summary, the embodiments of the present application determine the first vehicle state judgment result by detecting the vehicle state of the vehicle state detection model, combine the first vehicle state judgment result obtained online with the first preset weight to obtain the first state, and combine the second vehicle judgment result obtained offline in advance with the second preset weight to obtain the second state, to guide the online acquisition of the first state according to the second state obtained offline, thereby realizing the migration of offline learning ability to online learning ability, improving the learning ability of online learning, realizing the collaborative evolution of online learning system and offline learning system, and making both more intelligent.

[0057] The automatic driving data mining device provided by the present application is described below, and the automatic driving data mining device described below can be correspondingly referred to the automatic driving data mining method described above.

[0058] Figure 3 A structural schematic diagram of an automatic driving data mining device is shown, the device comprises:

[0059] The data acquisition module 31 acquires automatic driving data.

[0060] The state detection module 32 inputs the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training vehicle training data and a vehicle state true value corresponding to the vehicle training data.

[0061] The state judgment module 33 judges whether the vehicle state changes according to the vehicle predicted state and the previously acquired current vehicle state to obtain a first vehicle state switching result.

[0062] The data return module 34 compares the first vehicle state switching result with a second vehicle state switching result acquired offline based on the automatic driving data to obtain a first comparison result, determines whether the first comparison result meets a first preset condition, and determines to return the automatic driving data based on the determination that the first comparison result does not meet the first preset condition.

[0063] In the embodiment, the data acquisition module 31 includes: a data acquisition unit that acquires driving images and vehicle state data including vehicle steering and acceleration / deceleration information, and / or acquires radar point cloud data and vehicle state data including vehicle steering and acceleration / deceleration information, and / or acquires combined navigation data and vehicle state data including vehicle steering and acceleration / deceleration information.

[0064] Specifically, the data acquisition unit includes: an image acquisition subunit that acquires driving images in a target time period based on a camera; or a video acquisition subunit that acquires a video stream of a target region; and the image acquisition subunit extracts driving images in the target time period based on the video stream.

[0065] It should be noted that the image acquisition subunit includes: based on the video stream, a certain number of driving images before and after the current frame of driving image are collected, for example, ten frames of driving images before and after the current frame of driving image are collected, so as to avoid the case that the judgment result and the detection result based on a single frame of driving image are poor, and the specific number of driving images acquired can be set according to actual detection needs, which is not limited further herein.

[0066] The state detection module 32 includes: a data input unit that inputs the automatic driving data into a vehicle state detection model; a state detection module that performs state detection on the automatic driving data by using the vehicle state detection model to obtain a vehicle predicted state; and a data output unit that outputs the vehicle predicted state obtained by the vehicle state detection model.

[0067] In an optional embodiment, the apparatus further comprises a training module configured to train the vehicle state detection model before inputting the automatic driving data into the vehicle state detection model. Specifically, the training module comprises a vehicle training data acquisition unit configured to acquire vehicle training data and a vehicle state true value corresponding to the vehicle training data; and a training unit configured to train the to-be-trained model by taking the vehicle training data as input data for training and taking the vehicle state true value as a label for training, to obtain the vehicle state detection model for detecting the vehicle state.

[0068] The state judgment module 33 comprises a state judgment unit configured to judge whether the vehicle driving state is stable according to the vehicle predicted state and the previously acquired current vehicle state; if the vehicle does not appear to draw a dragon or an unexpected steering situation, it is determined that the vehicle driving state is stable; if the vehicle appears to draw a dragon or an unexpected steering situation, it is determined that the vehicle driving state is unstable. It should be noted that drawing a dragon means that the automatic driving vehicle will swing back and forth on the road.

[0069] In an optional embodiment, the apparatus further comprises an offline state judgment module configured to judge whether the vehicle driving state is stable according to the automatic driving data and in combination with prior experience or historical driving data before determining whether the first preset condition is met according to the first vehicle state switching result and the first preset weight and the second vehicle state switching result and the second preset weight which are previously acquired based on the automatic driving data offline.

[0070] In an optional embodiment, the apparatus further comprises a driving mode determination module configured to obtain a first vehicle driving mode switching decision according to the vehicle predicted state and the previously acquired current driving mode after obtaining the vehicle predicted state output by the vehicle state detection model; and a driving mode decision module configured to compare the first vehicle driving mode switching decision with a second vehicle driving mode switching decision which is previously acquired based on the automatic driving data offline, to obtain a second comparison result, to determine whether the second comparison result meets a second preset condition, and to determine to return the automatic driving data based on the determination that the second comparison result does not meet the second preset condition.

[0071] In addition, the apparatus further comprises an offline driving mode determination module configured to determine whether to switch the current driving mode according to the automatic driving data and in combination with prior experience or historical driving data before determining whether to switch the driving mode according to the vehicle predicted state and the previously acquired current driving mode, to obtain a second vehicle driving mode switching decision.

[0072] Further, the driving mode determination module is configured to determine to switch the current driving mode according to the first vehicle driving mode switching decision after obtaining the first vehicle driving mode switching decision; and the data feedback module 35 is configured to determine to feedback the automatic driving data based on that the driver takes over when the current driving mode is switched; and determine to feedback the automatic driving data based on that the driver does not take over when the current driving mode is switched, and if it is determined that the switched driving mode does not meet the expectation after the driving mode is switched.

[0073] Further, the driving mode determination module comprises: a vehicle state determination unit configured to determine that the vehicle driving state is stable according to the vehicle predicted state; and a mode switching unit configured to switch the current vehicle driving mode to the automatic driving Full AD mode based on that the vehicle control behavior determined based on the high-precision map is consistent with the driving state in the current LCC mode (no unintended steering or drawing a dragon will occur when switched to the Full AD mode) in a case that the current vehicle driving mode is the lane center keeping LCC mode; and switch the current vehicle driving mode to the LCC mode based on that the vehicle control behavior determined based on the high-precision map is inconsistent with the driving state in the current Full AD mode in a case that the current vehicle driving mode is the Full AD mode.

[0074] In addition, the vehicle state determination unit is further configured to determine that the vehicle driving state is unstable according to the vehicle predicted state; and the mode switching unit is further configured to switch the current vehicle driving mode to the fallback mode.

[0075] In an optional embodiment, the apparatus further comprises: a feedback data decision module configured to determine whether to feedback data according to the vehicle predicted state, the preset feedback data scene and the preset feedback rule after obtaining the vehicle predicted state output by the vehicle state detection model, to obtain a first feedback data decision; and the feedback data decision module is configured to determine whether to meet a third preset condition according to the first feedback data decision and a fifth preset weight, and a second feedback data decision based on the automatic driving data offline and a sixth preset weight in advance, and determine to feedback the automatic driving data based on that the third preset condition is not met.

[0076] In an optional embodiment, the apparatus further comprises: an offline driving mode determination module configured to determine whether to feedback data according to the automatic driving data and combine priori experience or historical driving data to obtain a second feedback data decision before determining whether to feedback data according to the vehicle predicted state, the preset feedback data scene and the preset feedback rule.

[0077] To sum up, the embodiment of the present application uses the vehicle state detection model to detect the vehicle state through the state detection module, determines the first vehicle state judgment result through the state judgment module, combines the first vehicle state judgment result obtained online with the first preset weight through the state acquisition module to obtain the first state, combines the second vehicle judgment result obtained offline with the second preset weight to obtain the second state, and guides the acquisition of the first state online according to the second state obtained offline, thereby realizing the migration of the offline learning ability to the online learning ability, improving the learning ability of online learning, realizing the collaborative evolution of the online learning system and the offline learning system, and making both more intelligent.

[0078] Figure 4 An example of an entity structure diagram of an electronic device is shown in Figure 4 As shown, the electronic device can include a processor 41, a communications interface 42, a memory 43, and a communications bus 44, wherein the processor 41, the communications interface 42, and the memory 43 communicate with each other through the communications bus 44. The processor 41 can call the logical instructions in the memory 43 to execute the automatic driving data mining method, which includes: obtaining automatic driving data; inputting the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training vehicle training data and vehicle state true values corresponding to the vehicle training data; determining whether the vehicle state changes according to the vehicle predicted state and the previously obtained current vehicle state to obtain a corresponding first vehicle state switching result; comparing the first vehicle state switching result with a second vehicle state switching result obtained offline based on the automatic driving data to obtain a first comparison result, determining whether the first comparison result meets a first preset condition, and determining to return the automatic driving data based on the non-compliance.

[0079] In addition, the logic instructions in the memory 43 described above can be implemented in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the embodiments of the method of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0080] In another aspect, the present application also provides a computer program product, the computer program product comprising a computer program, the computer program being stored in a non-transitory computer readable storage medium, and the computer program being executable by a processor to cause a computer to execute the automatic driving data mining method provided by the above-mentioned methods, the method comprising: obtaining automatic driving data; inputting the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training vehicle training data and a vehicle state true value corresponding to the vehicle training data; determining whether the vehicle state changes according to the vehicle predicted state and a previously obtained current vehicle state to obtain a first vehicle state switching result; comparing the first vehicle state switching result with a second vehicle state switching result obtained offline based on the automatic driving data in advance to obtain a first comparison result, determining whether the first comparison result meets a first preset condition, and based on the determination that the first comparison result does not meet the first preset condition, determining to return the automatic driving data.

[0081] In another aspect, the present application also provides a computer program product, the computer program product comprising a computer program, the computer program being stored in a non-transitory computer readable storage medium, and the computer program being executable by a processor to cause a computer to execute the automatic driving data mining method provided by the above-mentioned methods, the method comprising: obtaining automatic driving data; inputting the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training vehicle training data and a vehicle state true value corresponding to the vehicle training data; determining whether the vehicle state changes according to the vehicle predicted state and a previously obtained current vehicle state to obtain a first vehicle state switching result; comparing the first vehicle state switching result with a second vehicle state switching result obtained offline based on the automatic driving data in advance to obtain a first comparison result, determining whether the first comparison result meets a first preset condition, and based on the determination that the first comparison result does not meet the first preset condition, determining to return the automatic driving data.

[0082] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0083] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary universal hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, and the computer software products can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0084] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An automatic driving data mining method, characterized by, The method comprises: acquiring automatic driving data; inputting the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; wherein the vehicle state detection model is obtained by training vehicle training data and a vehicle state true value corresponding to the vehicle training data; determining whether a vehicle state changes according to the vehicle predicted state and a previously acquired current vehicle state to obtain a first vehicle state switching result; comparing the first vehicle state switching result and a second vehicle state switching result previously acquired offline based on the automatic driving data to obtain a first comparison result, determining whether the first comparison result meets a first preset condition, and based on a determination that the first comparison result does not meet the first preset condition, determining to return the automatic driving data; the first preset condition is whether a difference between the first vehicle state switching result with a first preset weight and the second vehicle state switching result with a second preset weight is within a first preset range, and based on a determination that the difference is not within the first preset range, determining that the first comparison result does not meet the first preset condition; the first preset weight and the second preset weight are determined according to the learning ability of an actual vehicle state detection model. 2.The automatic driving data mining method of claim 1, wherein, After obtaining the vehicle predicted state output by the vehicle state detection model, the method comprises: obtaining a first vehicle driving mode switching decision according to the vehicle predicted state and a previously acquired current driving mode; comparing the first vehicle driving mode switching decision and a second vehicle driving mode switching decision previously acquired offline based on the automatic driving data to obtain a second comparison result, determining whether the second comparison result meets a second preset condition, and based on a determination that the second comparison result does not meet the second preset condition, determining to return the automatic driving data. 3.The automatic driving data mining method of claim 2, wherein, After obtaining the first vehicle driving mode switching decision, the method comprises: determining to switch the current driving mode according to the first vehicle driving mode switching decision; based on the driver taking over when switching the current driving mode, determining to return the automatic driving data; based on the driver not taking over when switching the current driving mode, after completing the driving mode switching, if it is determined that the switched driving mode does not meet an expectation, determining to return the automatic driving data. 4.The automatic driving data mining method of claim 2, wherein, The method of obtaining the first vehicle driving mode switching decision comprises: if it is determined that the vehicle driving state is stable according to the vehicle predicted state, then: in a case where the current vehicle driving mode is a lane center keeping LCC mode, switching the current vehicle driving mode to an automatic driving Full AD mode based on a vehicle control behavior determined based on a high-definition map being consistent with a driving state in the LCC mode; in a case where the current vehicle driving mode is the automatic driving Full AD mode, switching the current vehicle driving mode to the LCC mode based on a vehicle control behavior determined based on the high-definition map being inconsistent with a driving state in the automatic driving Full AD mode. 5.The automatic driving data mining method of claim 2, wherein, The method of obtaining the first vehicle driving mode switching decision further comprises: If it is determined, according to the vehicle predicted state, that the vehicle driving state is unstable, the current vehicle driving mode is switched to a fallback mode in which the automatic driving is exited.

6. The automatic driving data mining method of claim 1, wherein, After the vehicle predicted state output by the vehicle state detection model is obtained, the method further includes: According to the vehicle predicted state, a preset data return scenario, and a preset data return rule, it is determined whether to return data, and a first data return decision is obtained. According to the first data return decision and a fifth preset weight, and a second data return decision and a sixth preset weight that are obtained offline based on the automatic driving data, it is determined whether a third preset condition is met, and based on the third preset condition not being met, it is determined that the automatic driving data is returned. 7.The automatic driving data mining method of claim 1, wherein, The vehicle state detection model is trained, including: Vehicle training data and a vehicle state true value corresponding to the vehicle training data are obtained; The vehicle training data is used as input data for training, and the vehicle state true value is used as a label for training, and a model to be trained is trained to obtain a vehicle state detection model for detecting a vehicle state.

8. An automatic driving data mining device characterized by comprising: The method includes: A data acquisition module acquires automatic driving data; A state detection module inputs the automatic driving data into a vehicle state detection model to obtain a vehicle predicted state output by the vehicle state detection model; the vehicle state detection model is trained according to vehicle training data and a vehicle state true value corresponding to the vehicle training data; A state judgment module judges whether a vehicle state has changed according to the vehicle predicted state and a previously obtained current vehicle state, and obtains a first vehicle state switching result; A data return module compares the first vehicle state switching result and a second vehicle state switching result obtained offline based on the automatic driving data to obtain a first comparison result, determines whether the first comparison result meets a first preset condition, and based on the first comparison result not meeting the first preset condition, determines that the automatic driving data is returned; The first preset condition is whether a difference between the first vehicle state switching result with a first preset weight and the second vehicle state switching result with a second preset weight is within a first preset range, and based on the difference not being within the first preset range, it is determined that the first preset condition is not met; The first preset weight and the second preset weight are determined according to the learning ability of an actual vehicle state detection model.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the automatic driving data mining method according to any one of claims 1 to 7. 10.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the automatic driving data mining method according to any one of claims 1 to 7.

Images