Data mining-based control method, device, system, equipment and storage medium

Abstract

The present invention discloses a data mining-based regulation method, device, equipment and storage medium, the method comprising: obtaining an initial candidate trajectory data set based on an original regulation algorithm; determining the scene category of the target vehicle, and processing the initial candidate trajectory data set based on the scene category to obtain a candidate trajectory data set; obtaining the decision level of each candidate trajectory in the candidate trajectory data set, and forming a decision level sequence; screening the candidate trajectory data set based on the decision level sequence to obtain a target trajectory, and outputting a report based on the target trajectory. The technical solution of the present invention determines the target trajectory based on data mining, which greatly improves the confidence of the target trajectory compared to the existing original regulation algorithm.

Description

Technical Field

[0001] The present invention belongs to the field of autonomous driving technology, and in particular relates to a data mining-based regulation and control method, device, system, equipment and storage medium. Background Art

[0002] Autonomous driving trajectory planning is mainly used to determine the vehicle's driving route. For example, if an autonomous driving vehicle is in an environment with multiple obstacles around it and needs to turn left, trajectory planning is required to determine the subsequent driving route.

[0003] The input of trajectory planning includes topological maps, obstacles and predicted trajectories of obstacles, the status of traffic lights, and other information such as positioning, navigation, and vehicle status. The output of trajectory planning is a trajectory, which is a function of time to position, which means that the vehicle is required to be at a specific position at a specific time.

[0004] At present, when an autonomous vehicle is driving, it directly controls the actual trajectory of the vehicle by referring to the recommended predicted trajectory; however, there are the following problems:

[0005] 1) The amount of information used by the regulatory control algorithm itself is less than that of the real scene, resulting in accuracy problems and poor confidence;

[0006] 2) The control algorithm does not use the driver’s own behavior information as input, and the planned trajectory is not flexible enough, resulting in a poor experience. Summary of the invention

[0007] The present invention aims to solve one of the technical problems in the related art at least to a certain extent. To this end, one object of the present invention is to propose a data mining-based regulation and control method, device, system, equipment and storage medium.

[0008] In order to solve the above technical problems, the embodiments of the present invention provide the following technical solutions:

[0009] A regulatory control method based on data mining, comprising:

[0010] Based on the original regulation and control algorithm, the initial candidate trajectory data set is obtained;

[0011] Determine a scene category in which the target vehicle is located, and based on the scene category, process the initial candidate trajectory data set to obtain a candidate trajectory data set;

[0012] Obtaining a decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence;

[0013] The candidate trajectory data set is screened based on the decision level sequence to obtain a target trajectory, and a first report is output based on the target trajectory.

[0014] Optionally, determining the scene category of the target vehicle and processing the initial candidate trajectory data set based on the scene category to obtain the candidate trajectory data set includes:

[0015] If the scene category is the first category, the initial candidate trajectory data set is used as the first candidate trajectory data set; or

[0016] If the scene category is the second category, screening each of the initial candidate trajectories in the initial candidate trajectory data set based on a ranking algorithm to determine a screened candidate trajectory data set;

[0017] Based on a search algorithm or a scene determination model, a supplementary candidate trajectory data set is obtained;

[0018] Based on the supplementary candidate trajectory dataset and the screened candidate trajectory dataset, a second candidate trajectory dataset is acquired.

[0019] Before obtaining the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence, the method further includes:

[0020] Determining the number of real-time trajectories of the candidate trajectories in the candidate trajectory data set;

[0021] Comparing the real-time track number with a track number threshold;

[0022] If the number of real-time trajectories is greater than the trajectory number threshold, determining a target decision level based on the decision level sequence, and determining the target trajectory based on the target decision level;

[0023] If the number of real-time trajectories is less than or equal to the trajectory number threshold, then based on an iterative algorithm, the value of the real-time trajectory number is increased until the number of real-time trajectories is greater than the trajectory number threshold.

[0024] Optionally, the acquiring a decision level of each candidate trajectory in the first candidate trajectory data set and forming a decision level sequence includes:

[0025] calculating each of the first candidate trajectories in the first candidate trajectory data set based on the ranking algorithm, obtaining a first decision level for each of the first candidate trajectories, and forming a first decision level sequence; or

[0026] Each second candidate trajectory of the second candidate trajectory data set is calculated based on the ranking algorithm to obtain a second decision level for each second candidate trajectory and form a second decision level sequence.

[0027] Optionally, also include:

[0028] Get the original data;

[0029] Determining the amount of data included in the original data;

[0030] If the data volume is greater than the data volume threshold, the original regulation algorithm is started;

[0031] If the data amount is less than or equal to the data amount threshold, inputting the real-time data of the target vehicle into the scene determination model;

[0032] Based on the scene determination model, outputting a target scene of the target vehicle;

[0033] Based on the target scene, obtaining a third candidate trajectory data set;

[0034] The third candidate trajectory data set is screened to obtain the target trajectory.

[0035] Optionally, also include:

[0036] Obtain historical target trajectory data sets and historical actual trajectory data sets;

[0037] Splitting the historical target trajectory data set and the historical actual trajectory data set to obtain a trajectory segment data set;

[0038] Calculating the trajectory segment data set based on an evaluation algorithm to obtain a calculation result;

[0039] Based on the calculation result, the parameters of the ranking algorithm are adjusted to obtain an adjusted ranking algorithm.

[0040] An embodiment of the present invention further provides a data mining-based regulation and control device, comprising:

[0041] An acquisition module is used to acquire an initial candidate trajectory data set based on an original regulation algorithm;

[0042] A determination module, used to determine the scene category where the target vehicle is located, and based on the scene category, process the initial candidate trajectory data set to obtain a candidate trajectory data set;

[0043] A calculation module, used for obtaining the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence;

[0044] A screening module is used to screen the candidate trajectory data set based on the decision level sequence to obtain a target trajectory, and output a first report based on the target trajectory.

[0045] An embodiment of the present invention further provides a regulation and control system based on data mining, comprising:

[0046] A judgment unit, used to determine whether to start the original regulation and control algorithm; wherein the original regulation and control algorithm is used to obtain an initial candidate trajectory data set;

[0047] A scene unit, used to determine the scene category of the target vehicle and obtain a candidate trajectory data set for each category of the scene;

[0048] A decision optimization unit, used to obtain the candidate trajectory data set, and calculate each candidate trajectory in the candidate trajectory data set, determine the decision level of each candidate trajectory, and form a decision level sequence; then screen the candidate trajectory data based on the decision level sequence to determine the target trajectory;

[0049] The adjustment unit is used to adjust the parameters of the decision optimization unit and the scenario unit, obtain the adjustment parameters, and output a second report based on the adjustment parameters.

[0050] An embodiment of the present invention further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method as described above when executing the computer program.

[0051] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method described above.

[0052] The embodiments of the present invention have the following technical effects:

[0053] The above technical solution of the present invention, 1) determines the target trajectory based on data mining, specifically obtains an initial candidate trajectory data set through an original regulation and control algorithm, obtains real-time status data of the target vehicle based on perception results and technical means such as high-precision maps, and supplements the second candidate trajectory data set based on the real-time status data, or obtains a third candidate trajectory data set based on the real-time status data. After obtaining the first candidate trajectory data set, the second candidate trajectory data set or the third candidate trajectory data set, the target trajectory is determined based on a hierarchical algorithm, and based on the feasible target trajectory after verification and the optimal hyperparameters (such as weight values) of the regulation and control algorithm, an optimization report of the regulation and control method is output; compared with the existing original regulation and control algorithm, the confidence of the target trajectory is greatly improved.

[0054] 2) A scene unit is set up to identify and confirm the scene in which the target vehicle is currently located. Different methods are used to obtain candidate trajectory data sets according to the different scene categories in which the target vehicle is located, thereby improving the confidence of the candidate trajectory data sets and thus improving the confidence of the target trajectory.

[0055] 3) The regulatory control system based on data mining provides targeted modification suggestions for the original regulatory control system and algorithm through big data to improve the safety, comfort and reliability of the system.

[0056] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a schematic diagram of the structure of a regulation and control system based on data mining provided by an embodiment of the present invention;

[0058] Figure 2 is a flow chart of a regulation and control algorithm based on data mining provided by an embodiment of the present invention;

[0059] Figure 3 This is an example of a process of a regulation algorithm based on data mining provided by an embodiment of the present invention;

[0060] Figure 4 This is another example of the process of the regulation algorithm based on data mining provided by an embodiment of the present invention;

[0061] Figure 5 This is a structural example of a regulation and control device based on data mining provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0063] In order to facilitate the understanding of the embodiments by those skilled in the art, some terms are explained:

[0064] (1) Cut-in: refers to the scene where a target vehicle cuts in.

[0065] (2) Cut-off: indicates the scene where the target vehicle is cut out.

[0066] (3)Cost function: cost function, cost function.

[0067] (4) Transformer model: a deep learning model.

[0068] like Figure 1 As shown, an embodiment of the present invention provides a regulation and control system based on data mining, including:

[0069] A judgment unit, used to determine whether to start the original regulation and control algorithm; wherein the original regulation and control algorithm is used to obtain an initial candidate trajectory data set;

[0070] Specifically, the judgment unit judges the data volume of the original data in the current database of the target vehicle based on the data volume threshold. If the data volume is greater than the data volume threshold, the judgment unit starts the original regulation and control algorithm. After the original regulation and control algorithm is started, the current state data of the target vehicle is obtained, and the current state data of the target vehicle is processed to output an initial candidate trajectory data set.

[0071] An optional embodiment of the present invention further includes a cold start unit, and a scene determination model is provided inside the cold start unit; specifically, if the data amount is less than or equal to the data amount threshold, the real-time status data of the target vehicle is sent to the scene determination model, and then the scene determination model processes the real-time status data of the target vehicle and outputs a third candidate trajectory data set.

[0072] The scene unit is used to determine the scene category of the target vehicle and obtain the candidate trajectory data set for each category of the scene; wherein the scene includes a first category of scenes and a second category of scenes; the scene unit outputs a first candidate trajectory data set in the first category of scenes; and the scene unit outputs a second candidate trajectory data set in the second category of scenes.

[0073] A decision optimization unit, used to obtain the candidate trajectory data set, and calculate each candidate trajectory in the candidate trajectory data set, determine the decision level of each candidate trajectory, and form a decision level sequence; then screen the candidate trajectory data based on the decision level sequence to determine the target trajectory;

[0074] Specifically, the decision optimization unit obtains the first candidate trajectory data set, the second candidate trajectory data set, or the third candidate trajectory data set, and after determining the target trajectory, forms a report based on the target trajectory and outputs the report.

[0075] Among them, the decision optimization unit interacts with the judgment unit, the cold start unit and the scenario unit based on the network.

[0076] The adjustment unit is used to adjust the parameters of the decision optimization unit and the scenario unit, obtain the adjustment parameters, and output a second report based on the adjustment parameters.

[0077] In an embodiment of the present invention, an original regulation and control system is provided in the target vehicle, and the original regulation and control system implements data interaction with the judgment unit and the scene unit based on the network, so as to implement the original regulation and control system to send the calculated initial candidate trajectory data set to the judgment unit and the scene unit based on the network.

[0078] Furthermore, the original regulation and control system is provided with an original regulation and control algorithm unit; wherein the original regulation and control algorithm unit calculates the real-time state data of the target vehicle based on the original regulation and control algorithm to obtain an initial candidate trajectory data set.

[0079] In an optional embodiment of the present invention, the decision optimization unit may be connected to a display screen, and after the report is generated, the report is sent to the display screen for display.

[0080] In an embodiment of the present invention, a scene unit is provided to identify and confirm the scene in which the target vehicle is currently located. Different methods for obtaining candidate trajectory data sets are used according to different scene categories in which the target vehicle is located, thereby improving the confidence of the candidate trajectory data sets and further improving the confidence of the target trajectory.

[0081] Furthermore, an embodiment of the present invention further provides an adjustment unit, which is used to adjust some operating parameters of the system to reduce system errors and further improve the confidence of the candidate trajectory data set.

[0082] The embodiments of the present invention are based on a data mining regulatory control system, which provides targeted modification suggestions for the original regulatory control system and algorithm by means of big data, thereby improving the safety, comfort and reliability of the system.

[0083] like Figure 2 As shown, an embodiment of the present invention provides a regulation and control method based on data mining, which is applied to the above system, including:

[0084] Step S1: Based on the original regulation algorithm, obtain the initial candidate trajectory data set;

[0085] Specifically, the real-time status data of the target vehicle can be obtained based on the system's perception result data and high-precision map information. The original regulation and control algorithm can be accessed based on the white box, and then the real-time status data of the target vehicle can be calculated based on the original regulation and control algorithm to obtain the initial candidate trajectory data set.

[0086] For example, based on the original control algorithm, the lateral trajectory beam and the longitudinal trajectory beam are obtained respectively, and the trajectory is most basically generated by a polynomial method;

[0087] For example: s = c0 + c1t + c1t 2 +c1t 3 +c1t 4 ;

[0088] In the formula, s represents the position; c0 represents the hyperparameter; t represents the time.

[0089] In actual application scenarios, the system's input data includes not only the above-mentioned initial candidate trajectory data set, but also the system's perception result data and high-precision map information.

[0090] It should be noted that the original regulation and control algorithm can be implemented based on the existing algorithm steps, so the embodiments of the present invention will not elaborate on this.

[0091] Step S2: determining the scene category of the target vehicle, and based on the scene category, processing the initial candidate trajectory data set to obtain a candidate trajectory data set;

[0092] In an embodiment of the present invention, an initial model is constructed, and the initial scene determination model is trained multiple times to obtain a scene determination model; wherein the scene determination model is used to output vector representation data of the scene in which the target vehicle is currently located based on the real-time state data of the target vehicle; for example, the initial model may be a transformer model;

[0093] In actual application scenarios, first, obtain the initial model training set, specifically:

[0094] Based on the database, historical state data of the target vehicle is obtained, including environment and scene data; wherein the form of the environment and scene of each frame is defined as a situation, and every 30 frames of situation is defined as a scene;

[0095] Specifically, the situation can be defined based on the feature angle. For example, the situation is divided into multiple coordinate points by dividing the situation into a nine-square grid with the target vehicle as the center and the lane lines as the divisions. Then, the obstacle conditions, road characteristics, and weather characteristics of each trajectory coordinate point are defined. Then, the defined situation is sorted according to the time series to form a scene sequence based on the time series, and the initial model training set is obtained based on the scene sequence based on the time series.

[0096] Input the initial model training set into the initial model, such as the transformer model, convert the coordinate point information into vector data based on the transformer model, and then output the vector representation data of the scene; repeat the above steps to train and obtain the scene determination model (for example: the target transformer model).

[0097] Specifically, determining the scene category of the target vehicle and processing the initial candidate trajectory data set based on the scene category to obtain the candidate trajectory data set includes:

[0098] If the scene category is the first category, the initial candidate trajectory data set is used as the first candidate trajectory data set; or

[0099] If the scene category is the second category, screening each of the initial candidate trajectories in the initial candidate trajectory data set based on a ranking algorithm to determine a screened candidate trajectory data set;

[0100] Based on a search algorithm or a scene determination model, a supplementary candidate trajectory data set is obtained;

[0101] Based on the supplementary candidate trajectory dataset and the screened candidate trajectory dataset, a second candidate trajectory dataset is acquired.

[0102] In an embodiment of the present invention, after obtaining the initial candidate trajectory data set, the scene category in which the target vehicle is currently located is first determined; wherein the scene category includes a first category scene and a second category scene;

[0103] Specifically, the first type of scenarios may include multiple common driving scenarios; for example, weather conditions, overtaking, lane changes, and certain traffic conditions;

[0104] The second type of scenes may include cut-in and cut-off scenes.

[0105] If it is determined after judgment that the target vehicle is currently in the first type of scenario, for example, lane change, the acquired initial candidate trajectory data set is directly used as the first candidate trajectory data set, and then the first candidate trajectory data set is output.

[0106] If it is determined that the target vehicle is currently in the second type of scenario, for example, another vehicle suddenly cuts in front of the target vehicle, the confidence of the initial candidate trajectory data set will be reduced, for example, safety or comfort will be reduced.

[0107] In order to solve the problem of reduced confidence in the initial candidate trajectory dataset, an embodiment of the present invention performs screening processing on the initial candidate trajectory dataset.

[0108] Specifically, a screening threshold is preset, and an initial decision level of each initial candidate trajectory in the initial candidate trajectory data set is calculated based on a level algorithm;

[0109] Among them, the hierarchical algorithm can be a cost function, which calculates the comfort cost, collision cost, centripetal acceleration cost, lateral offset cost and lateral comfort cost of each initial candidate trajectory; then, the cost function follows the weight parameters preset by the hierarchical algorithm, that is, each category of cost corresponds to a different weight, for example: the weight corresponding to the comfort cost is 0.1, the weight corresponding to the collision cost is 0.3, the weight corresponding to the centripetal acceleration cost is 0.2, the weight corresponding to the lateral offset cost is 0.15 and the weight corresponding to the lateral comfort cost is 0.25, and the comfort cost, collision cost, centripetal acceleration cost, lateral offset cost and lateral comfort cost are weighted and summed, and finally the initial decision level of each initial candidate trajectory, that is, the calculation result of the cost function, is calculated.

[0110] Then, the initial decision levels corresponding to all the initial candidate trajectories in the initial candidate trajectory data set are compared with the screening threshold. If the initial decision level is greater than the screening threshold, the initial candidate trajectory is retained; if the initial decision level is less than or equal to the screening threshold, the initial candidate trajectory is deleted; finally, based on all the retained initial candidate trajectories, a screening candidate trajectory data set is obtained.

[0111] Furthermore, in an embodiment of the present invention, after obtaining the screening candidate trajectory data set, the driving routes of human drivers of the same driving type (e.g., in certain similar scene states) are retrieved from the database of the target vehicle;

[0112] For example: determine the current scene state of the target vehicle: for example: determine that the current scene state of the target vehicle is a bridge and the weather is snowing, determine the current scene state of the target vehicle based on these two states, and then search the database based on the "snowy bridge", find the historical scene state corresponding to the scene state, obtain the similarity between the scene state and the historical scene state, and compare the similarity with the scene similarity threshold. If the similarity is greater than the scene similarity threshold, use the scene state as the output result;

[0113] If there are multiple historical scenes whose similarity is greater than the scene similarity threshold, the similarities corresponding to these historical scenes are sorted, and the historical scene at the front after the similarity sorting is output as the search result;

[0114] Then, based on the historical scene, the database is searched again to obtain the historical trajectory data set corresponding to the historical scene;

[0115] The historical trajectory dataset is used as a supplementary candidate trajectory dataset, and then the supplementary candidate trajectory dataset and the screened candidate dataset are output as a second candidate trajectory dataset.

[0116] Furthermore, if, based on the scene category in which the target vehicle is currently located, after searching the database of the target vehicle, no scene similar to the scene currently described by the target vehicle is found, that is, the similarity between the historical scenes in the database and the scene currently described by the target vehicle is less than or equal to the scene similarity threshold, then the real-time state data information of the target vehicle is obtained, and based on the real-time state data information, combined with the above algorithm, a real-time scene sequence based on time series is generated, the real-time scene sequence is input into the scene determination model, and then the vector representation data of the current scene of the target vehicle is output;

[0117] Based on the vector representation data, the database is searched, the vector representation data of the scene where the target vehicle is currently located is compared with the vector data of the historical scene in the database, and a comparison result is obtained; then the comparison result is compared with the scene similarity threshold, and if the comparison result is greater than the scene similarity threshold, the historical scene corresponding to the comparison result is output as the search result;

[0118] Then, based on the historical scene, the database is searched again to obtain the historical trajectory data set corresponding to the historical scene;

[0119] The historical trajectory dataset is used as a supplementary candidate trajectory dataset, and then the supplementary candidate trajectory dataset and the screened candidate dataset are output as a second candidate trajectory dataset.

[0120] Step S3: obtaining the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence;

[0121] Specifically, before obtaining the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence, the method further includes:

[0122] Determining the number of real-time trajectories of the candidate trajectories in the candidate trajectory data set;

[0123] Comparing the real-time track number with a track number threshold;

[0124] If the number of real-time trajectories is greater than the trajectory number threshold, determining a target decision level based on the decision level sequence, and determining the target trajectory based on the target decision level;

[0125] If the number of real-time trajectories is less than or equal to the trajectory number threshold, then based on an iterative algorithm, the value of the real-time trajectory number is increased until the number of real-time trajectories is greater than the trajectory number threshold.

[0126] In the embodiment of the present invention, in order to ensure that the confidence of the target trajectory finally obtained meets the requirements, before determining the decision level of each candidate trajectory level, the confidence of the candidate trajectory data set is first confirmed;

[0127] Specifically, a trajectory number threshold is set, for example, 100, and each time the confidence confirmation of the first candidate trajectory data set or the second candidate trajectory data set is performed, the total number of trajectories of the first candidate trajectory data set or the second candidate trajectory data set is obtained;

[0128] Taking the first candidate trajectory data set as an example, if the number of candidate trajectories included in the first candidate trajectory data set is 80, that is, the number of real-time trajectories is less than the trajectory number threshold, the first candidate trajectory data set is stored, and the original regulation algorithm is returned to continue to obtain the initial candidate trajectory data set, and then the above steps are repeated based on the iterative algorithm. Specifically, the first supplementary candidate trajectory data set of the first candidate trajectory data set is added to the first candidate trajectory data set to obtain the number of real-time trajectories, and the updated number of real-time trajectories is obtained. The updated number of real-time trajectories is compared with the trajectory number threshold. If the updated number of real-time trajectories is greater than the trajectory number threshold, the iteration process is stopped;

[0129] If the updated real-time trajectory number is still less than or equal to the trajectory number threshold, return to the original regulation and control algorithm to continue to obtain the initial candidate trajectory data set;

[0130] …

[0131] The iteration process stops when the number of real-time trajectories exceeds the trajectory threshold.

[0132] After obtaining the first candidate trajectory data set or the second candidate trajectory data set that meets the confidence level, the target trajectory is determined based on the first candidate trajectory data set or the second candidate trajectory data set.

[0133] Specifically, obtaining the decision level of each candidate trajectory in the first candidate trajectory data set and forming a decision level sequence includes:

[0134] calculating each of the first candidate trajectories in the first candidate trajectory data set based on the ranking algorithm, obtaining a first decision level for each of the first candidate trajectories, and forming a first decision level sequence; or

[0135] Each second candidate trajectory of the second candidate trajectory data set is calculated based on the ranking algorithm to obtain a second decision level for each second candidate trajectory and form a second decision level sequence.

[0136] In an embodiment of the present invention, the values ​​of comfort cost, collision cost, centripetal acceleration cost, lateral offset cost and lateral comfort cost of each candidate trajectory in the first candidate trajectory data set or the second candidate trajectory data set are calculated, and preset weight values ​​corresponding to the values ​​of comfort cost, collision cost, centripetal acceleration cost, lateral offset cost and lateral comfort cost are obtained, and then the weighted sum of the above performance costs is calculated, that is, the first decision level or the second decision level of each candidate trajectory is obtained;

[0137] Arrange the plurality of first decision levels or the plurality of second decision levels according to numerical values, for example, in order from large to small, to obtain a first decision level sequence or a second decision level sequence.

[0138] Step S4: Screening the candidate trajectory data set based on the decision level sequence to obtain the target trajectory, and outputting a first report based on the target trajectory. In an embodiment of the present invention, after obtaining the first decision level sequence to obtain the second decision level sequence, taking the second decision level sequence as an example:

[0139] Arranging all candidate trajectory data in the second candidate trajectory data set in the order of the corresponding second decision level sequence, a trajectory coordinate sequence can be obtained;

[0140] The track points are combined with the properties of the target vehicle to form a target vehicle area, that is, the target vehicle is combined with the track points in the form of a vehicle, so as to confirm the positional relationship between the target vehicle and the track points; similarly, the obstacle area is obtained to determine whether there is an intersection between the target vehicle area and the obstacle area;

[0141] If the target vehicle area and the obstacle area do not intersect, the candidate trajectory that is at the front in the second decision level sequence, i.e., the candidate trajectory corresponding to the highest decision level, is the target trajectory;

[0142] If the target vehicle area and the obstacle area intersect, the candidate trajectories corresponding to the last decision level in the second decision level sequence are deleted;

[0143] Among them, a decision level threshold may be set, for example, set to 10%, that is, the last 10% of the decision levels in the second decision level sequence are deleted;

[0144] Then, the number of remaining trajectories of the candidate trajectory data corresponding to the remaining 90% of the second decision level sequence is compared with the trajectory number threshold. If the number of remaining trajectories is greater than the trajectory number threshold, the candidate trajectory at the front position in the second level sequence, that is, the candidate trajectory corresponding to the highest decision level, is the target trajectory.

[0145] If the number of remaining trajectories is less than or equal to the trajectory number threshold, the remaining second candidate trajectory data set is stored, and the original control algorithm is returned to continue to obtain the initial candidate trajectory data set, and then the above steps are repeated based on the iterative algorithm to add the number of real-time trajectories of the first supplementary candidate trajectory data set of the second candidate trajectory data set and the second candidate trajectory data set to obtain an updated number of real-time trajectories;

[0146] …

[0147] The iteration process stops when the number of real-time trajectories exceeds the trajectory threshold.

[0148] Repeat the above steps to determine the target trajectory based on the hierarchical algorithm, and output a first report based on the target trajectory.

[0149] An optional embodiment of the present invention further includes:

[0150] Get the original data;

[0151] Determining the amount of data included in the original data;

[0152] If the data volume is greater than the data volume threshold, the original regulation algorithm is started;

[0153] If the data amount is less than or equal to the data amount threshold, inputting the real-time data of the target vehicle into the scene determination model;

[0154] Based on the scene determination model, outputting a target scene of the target vehicle;

[0155] Based on the target scene, obtaining a third candidate trajectory data set;

[0156] The third candidate trajectory data set is screened to obtain the target trajectory.

[0157] In the embodiment of the present invention, during the initial use of the target vehicle, the amount of data in the database may be relatively small. However, as the mileage of the target vehicle continues to increase, the amount of data in the database will gradually increase.

[0158] In actual application scenarios, before determining the target trajectory, the data volume of the original data in the database is first judged based on the data volume threshold; when the data volume is less than the data volume threshold, if the initial candidate trajectory data set is obtained based on the original regulation algorithm, the confidence of the initial candidate trajectory data set obtained at this time is low. Therefore, in an embodiment of the present invention, when the data volume is less than the data volume threshold, the original regulation algorithm is not started, or the initial trajectory candidate trajectory data set output by the original regulation algorithm is ignored.

[0159] In actual application scenarios, based on the real-time status data of the target vehicle, a real-time scene sequence based on time series is obtained, and then the real-time scene sequence is input into the scene determination model to output the vector representation data of the real-time scene;

[0160] Search the database based on the vector representation data, and after finding a target scene similar to the vector representation data, determine whether the found target scene meets the requirements based on the scene similarity threshold. If it meets the requirements, the target scene is used as the search result and output;

[0161] Then, based on the output target scene, the data is searched again to obtain a third candidate trajectory data set matching the target scene;

[0162] The number of trajectories in the third candidate trajectory data set is obtained, and the third candidate trajectory data set is judged based on the trajectory number threshold; for example, if the trajectory number threshold is 50 and the number of candidate trajectories included in the third candidate trajectory data set is 60, the decision level of each candidate trajectory is determined based on the level algorithm, and then a third decision level sequence is obtained;

[0163] Similarly, based on the evaluation algorithm involved in the embodiment of the present invention, which will not be elaborated here, it can be determined that when the data volume of the original data is less than the data volume threshold, a target trajectory with a higher confidence level is determined.

[0164] Generally, the threshold of the number of trajectories is smaller than the threshold of the number of trajectories.

[0165] like Figure 3 As shown, an optional embodiment of the present invention further includes:

[0166] Step S31: Acquire a historical target trajectory data set and a historical actual trajectory data set;

[0167] Step S32: splitting the historical target trajectory data set and the historical actual trajectory data set to obtain a trajectory segment data set;

[0168] Step S33: calculating the trajectory segment data set based on the evaluation algorithm to obtain a calculation result;

[0169] Step S34: Based on the calculation result, the parameters of the grading algorithm are adjusted to obtain an adjusted grading algorithm.

[0170] Since the target vehicle may have systematic errors during use, the confidence of the target trajectory may be affected.

[0171] In order to solve the above technical problems, an embodiment of the present invention obtains a historical target trajectory data set and a historical actual trajectory data set at regular intervals, such as 2 days, 5 days or 10 days, etc.; wherein the number of the historical target trajectory data set and the historical actual data set obtained can be determined according to actual needs;

[0172] The evaluation algorithm may specifically select KL divergence. If the KL divergence value of the trajectory segment data set is large, it indicates that the historical target trajectory data set is greatly different from the historical actual trajectory data set, which will affect the driving comfort. Therefore, the weight value corresponding to the current comfort cost is increased.

[0173] For example: adjust the weight value of comfort cost from 0.1 to 0.15, etc.

[0174] By analogy, the evaluation algorithm can also select other reference data to adjust the weight values ​​corresponding to other performance costs;

[0175] Furthermore, the weights may be adjusted in combination with sexual function indicators; for example, overtaking success rate, etc.

[0176] In an embodiment of the present invention, after determining one or more adjusted weight values, the grading algorithm is adjusted based on the one or more adjusted weight values ​​to obtain an adjusted grading algorithm; and the target trajectory is determined based on the adjusted grading algorithm, thereby achieving the technical effect of ensuring the confidence of the target trajectory and even improving the confidence of the target trajectory.

[0177] In an embodiment of the present invention, a target trajectory is determined based on data mining. Specifically, an initial candidate trajectory data set is obtained through an original regulation and control algorithm. Real-time status data of the target vehicle is obtained based on perception results and technical means such as high-precision maps. A second candidate trajectory data set is supplemented based on the real-time status data, or a third candidate trajectory data set is obtained based on the real-time status data. After obtaining the first candidate trajectory data set, the second candidate trajectory data set or the third candidate trajectory data set, the target trajectory is determined based on a hierarchical algorithm. Based on the feasible target trajectory after verification and the optimal hyperparameters (such as weight values) of the regulation and control algorithm, a regulation and control method optimization report is output. Compared with the existing original regulation and control algorithm, the confidence of the target trajectory is greatly improved.

[0178] like Figure 4 As shown, the above embodiments of the present invention can be implemented based on the following implementation methods:

[0179] 1) The amount of raw data in the database of the target vehicle is judged; if the amount of raw data is less than the data amount threshold, the real-time status data of the target vehicle is obtained, and then a real-time scene sequence based on time series is obtained based on the real-time status data of the target vehicle, and the real-time scene sequence is input as input data to the scene determination model to output the target scene; based on the target scene, a third candidate trajectory data set is obtained; then the third candidate trajectory data set is judged based on the trajectory number threshold; if the number of candidate trajectories included in the third candidate trajectory data set is greater than the trajectory number threshold, the target trajectory is determined based on the hierarchical algorithm; if the number of candidate trajectories included in the third candidate trajectory data set is less than or equal to the trajectory number threshold, the number of candidate trajectories included in the third candidate trajectory data set is continuously increased based on the iterative algorithm until the number of candidate trajectories included in the supplemented third candidate data set is greater than the trajectory number threshold.

[0180] 2) If the data volume is greater than the data volume threshold, the original regulation algorithm is started to obtain the initial candidate trajectory data set;

[0181] 3) Determine the scene category in which the target vehicle is currently located;

[0182] 4) If the target vehicle is currently in the first type of scenario, such as lane change, the initial candidate trajectory data set is used as the first candidate trajectory data set; the first candidate trajectory data set is judged based on the trajectory number threshold, and if the number of real-time trajectories contained in the first candidate trajectory data set is greater than the trajectory number threshold, the target trajectory is determined based on the equivalent algorithm; and based on the feasible target trajectory after verification and the optimal hyperparameters of the control algorithm, a control method optimization report is output;

[0183] If the number of real-time trajectories included in the first candidate trajectory data set is less than or equal to the trajectory number threshold, the number of candidate trajectories included in the first candidate trajectory data set is continuously increased based on an iterative algorithm until the number of candidate trajectories included in the supplemented first candidate trajectory data set is greater than the trajectory number threshold.

[0184] If the scene the target vehicle is currently in is the second type of scene, the initial candidate trajectory data set is screened based on the screening algorithm to obtain a screened candidate trajectory data set;

[0185] 6) Obtain the current scene state of the target vehicle, and obtain a supplementary candidate trajectory data set based on the scene state; summarize the screened candidate trajectory data set and the supplementary candidate trajectory data set to obtain a second candidate trajectory data set; judge the second candidate trajectory data set based on the trajectory number threshold, and if the number of real-time trajectories contained in the second candidate trajectory data set is greater than the trajectory number threshold, determine the target trajectory based on the equivalent algorithm; and output a regulation and control method optimization report based on the feasible target trajectory after verification and the optimal hyperparameters of the regulation and control algorithm;

[0186] If the number of real-time trajectories included in the second candidate trajectory data set is less than or equal to the trajectory number threshold, the number of candidate trajectories included in the second candidate trajectory data set is continuously increased based on an iterative algorithm until the number of candidate trajectories included in the supplemented second candidate trajectory data set is greater than the trajectory number threshold.

[0187] like Figure 5 As shown, an embodiment of the present invention further provides a data mining-based regulation and control device 500, comprising:

[0188] An acquisition module 501 is used to acquire an initial candidate trajectory data set based on an original regulation algorithm;

[0189] A determination module 502 is used to determine the scene category where the target vehicle is located, and based on the scene category, process the initial candidate trajectory data set to obtain a candidate trajectory data set;

[0190] A calculation module 503 is used to obtain the decision level of each candidate trajectory in the candidate trajectory data set and form a decision level sequence;

[0191] The screening module 504 is configured to screen the candidate trajectory data set based on the decision level sequence to obtain a target trajectory, and output a first report based on the target trajectory.

[0192] Optionally, determining the scene category of the target vehicle and processing the initial candidate trajectory data set based on the scene category to obtain the candidate trajectory data set includes:

[0193] If the scene category is the first category, the initial candidate trajectory data set is used as the first candidate trajectory data set; or

[0194] If the scene category is the second category, screening each of the initial candidate trajectories in the initial candidate trajectory data set based on a ranking algorithm to determine a screened candidate trajectory data set;

[0195] Based on a search algorithm or a scene determination model, a supplementary candidate trajectory data set is obtained;

[0196] Based on the supplementary candidate trajectory dataset and the screened candidate trajectory dataset, a second candidate trajectory dataset is acquired.

[0197] Before obtaining the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence, the method further includes:

[0198] Determining the number of real-time trajectories of the candidate trajectories in the candidate trajectory data set;

[0199] Comparing the real-time track number with a track number threshold;

[0200] If the number of real-time trajectories is greater than the trajectory number threshold, determining a target decision level based on the decision level sequence, and determining the target trajectory based on the target decision level;

[0201] If the number of real-time trajectories is less than or equal to the trajectory number threshold, then based on an iterative algorithm, the value of the real-time trajectory number is increased until the number of real-time trajectories is greater than the trajectory number threshold.

[0202] Optionally, the acquiring a decision level of each candidate trajectory in the first candidate trajectory data set and forming a decision level sequence includes:

[0203] calculating each of the first candidate trajectories in the first candidate trajectory data set based on the ranking algorithm, obtaining a first decision level for each of the first candidate trajectories, and forming a first decision level sequence; or

[0204] Each second candidate trajectory of the second candidate trajectory data set is calculated based on the ranking algorithm to obtain a second decision level for each second candidate trajectory and form a second decision level sequence.

[0205] Optionally, also include:

[0206] Get the original data;

[0207] Determining the amount of data included in the original data;

[0208] If the data volume is greater than the data volume threshold, the original regulation algorithm is started;

[0209] If the data amount is less than or equal to the data amount threshold, inputting the real-time data of the target vehicle into the scene determination model;

[0210] Based on the scene determination model, outputting a target scene of the target vehicle;

[0211] Based on the target scene, obtaining a third candidate trajectory data set;

[0212] The third candidate trajectory data set is screened to obtain the target trajectory.

[0213] Optionally, also include:

[0214] Obtain historical target trajectory data sets and historical actual trajectory data sets;

[0215] Splitting the historical target trajectory data set and the historical actual trajectory data set to obtain a trajectory segment data set;

[0216] Calculating the trajectory segment data set based on an evaluation algorithm to obtain a calculation result;

[0217] Based on the calculation result, the parameters of the ranking algorithm are adjusted to obtain an adjusted ranking algorithm.

[0218] An embodiment of the present invention further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method as described above when executing the computer program.

[0219] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method described above.

[0220] In addition, other structures and functions of the device in the embodiment of the present invention are known to those skilled in the art and will not be described in detail here to reduce redundancy.

[0221] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0222] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0223] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0224] In the description of the present invention, it is to be understood that the terms “center”, “longitudinal”, “lateral”, “length”, “width”, “thickness”, “up”, “down”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inside”, “outside”, “clockwise”, “counterclockwise”, “axial”, “radial”, “circumferential”, etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0225] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0226] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0227] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being "above", "above" or "above" a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being "below", "below" or "below" a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0228] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A regulatory control method based on data mining, characterized in that: include: Based on the original regulation and control algorithm, the initial candidate trajectory data set is obtained; Determine a scene category in which the target vehicle is located, and based on the scene category, process the initial candidate trajectory data set to obtain a candidate trajectory data set; Obtaining a decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence; Screening the candidate trajectory data set based on the decision level sequence to obtain a target trajectory, and outputting a report based on the target trajectory; The determining the scene category of the target vehicle and processing the initial candidate trajectory data set based on the scene category to obtain the candidate trajectory data set includes: If the scene category is the first category, the initial candidate trajectory data set is used as the first candidate trajectory data set; or If the scene category is the second category, screening each of the initial candidate trajectories in the initial candidate trajectory data set based on a ranking algorithm to determine a screened candidate trajectory data set; Based on a search algorithm or a scene determination model, a supplementary candidate trajectory data set is obtained; Based on the supplementary candidate trajectory dataset and the screened candidate trajectory dataset, acquiring a second candidate trajectory dataset; The obtaining of the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence includes: calculating each of the first candidate trajectories in the first candidate trajectory data set based on the ranking algorithm, obtaining a first decision level for each of the first candidate trajectories, and forming a first decision level sequence; or Based on the grade algorithm, each second candidate trajectory of the second candidate trajectory data set is calculated to obtain a second decision grade for each second candidate trajectory, and a second decision grade sequence is formed; wherein the first type of scene includes a plurality of ordinary driving scenes, and the second type of scene includes scenes with a target vehicle cutting in and a target vehicle cutting out.

2. The method according to claim 1, characterized in that Before obtaining the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence, the method further includes: Determining the number of real-time trajectories of the candidate trajectories in the candidate trajectory data set; Comparing the real-time track number with a track number threshold; If the number of real-time trajectories is greater than the trajectory number threshold, determining a target decision level based on the decision level sequence, and determining the target trajectory based on the target decision level; If the number of real-time trajectories is less than or equal to the trajectory number threshold, then based on an iterative algorithm, the value of the real-time trajectory number is increased until the number of real-time trajectories is greater than the trajectory number threshold.

3. The method according to claim 1, characterized in that Also includes: Get the original data; Determining the amount of data included in the original data; If the data volume is greater than the data volume threshold, the original regulation algorithm is started; If the data amount is less than or equal to the data amount threshold, inputting the real-time data of the target vehicle into the scene determination model; Based on the scene determination model, outputting a target scene of the target vehicle; Based on the target scene, obtaining a third candidate trajectory data set; The third candidate trajectory data set is screened to obtain the target trajectory.

4. The method according to claim 1, characterized in that: Also includes: Obtain historical target trajectory data sets and historical actual trajectory data sets; Splitting the historical target trajectory data set and the historical actual trajectory data set to obtain a trajectory segment data set; Calculating the trajectory segment data set based on an evaluation algorithm to obtain a calculation result; Based on the calculation result, the parameters of the original regulation and control algorithm of the grade algorithm are adjusted to obtain an adjusted grade algorithm.

5. A data mining-based regulation and control device, characterized in that: include: An acquisition module is used to acquire an initial candidate trajectory data set based on an original regulation algorithm; A determination module, used to determine the scene category where the target vehicle is located, and based on the scene category, process the initial candidate trajectory data set to obtain a candidate trajectory data set; A calculation module, used for obtaining the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence; a screening module, configured to screen the candidate trajectory data set based on the decision level sequence to obtain a target trajectory, and output a first report based on the target trajectory; The determining the scene category of the target vehicle and processing the initial candidate trajectory data set based on the scene category to obtain the candidate trajectory data set includes: If the scene category is the first category, the initial candidate trajectory data set is used as the first candidate trajectory data set; or If the scene category is the second category, screening each of the initial candidate trajectories in the initial candidate trajectory data set based on a ranking algorithm to determine a screened candidate trajectory data set; Based on a search algorithm or a scene determination model, a supplementary candidate trajectory data set is obtained; Based on the supplementary candidate trajectory dataset and the screened candidate trajectory dataset, acquiring a second candidate trajectory dataset; The obtaining of the decision level of each candidate trajectory in the candidate trajectory data set and forming a decision level sequence includes: calculating each of the first candidate trajectories in the first candidate trajectory data set based on the ranking algorithm, obtaining a first decision level for each of the first candidate trajectories, and forming a first decision level sequence; or Based on the grade algorithm, each second candidate trajectory of the second candidate trajectory data set is calculated to obtain a second decision grade for each second candidate trajectory, and a second decision grade sequence is formed; wherein the first type of scene includes a plurality of ordinary driving scenes, and the second type of scene includes scenes with a target vehicle cutting in and a target vehicle cutting out.

6. A regulatory control system based on data mining, characterized in that: include: A judgment unit, used to determine whether to start the original regulation and control algorithm; wherein the original regulation and control algorithm is used to obtain an initial candidate trajectory data set; A scene unit, used to determine the scene category of the target vehicle and obtain a candidate trajectory data set for each category of the scene; A decision optimization unit, configured to obtain the candidate trajectory data set, calculate each candidate trajectory in the candidate trajectory data set, determine a decision level of each candidate trajectory, and form a decision level sequence; then screen the candidate trajectory data based on the decision level sequence, determine a target trajectory, and output a first report based on the target trajectory; An adjustment unit, configured to adjust parameters of the decision optimization unit and the scenario unit, obtain adjustment parameters, and output a second report based on the adjustment parameters; The step of determining the scene category of the target vehicle and obtaining a candidate trajectory data set for each scene category includes: If the scene category is the first category, the initial candidate trajectory data set is used as the first candidate trajectory data set; or If the scene category is the second category, screening each of the initial candidate trajectories in the initial candidate trajectory data set based on a ranking algorithm to determine a screened candidate trajectory data set; Based on a search algorithm or a scene determination model, a supplementary candidate trajectory data set is obtained; Based on the supplementary candidate trajectory dataset and the screened candidate trajectory dataset, acquiring a second candidate trajectory dataset; The step of obtaining the candidate trajectory data set, calculating each candidate trajectory in the candidate trajectory data set, determining a decision level for each candidate trajectory, and forming a decision level sequence includes: calculating each of the first candidate trajectories in the first candidate trajectory data set based on the ranking algorithm, obtaining a first decision level for each of the first candidate trajectories, and forming a first decision level sequence; or Based on the grade algorithm, each second candidate trajectory of the second candidate trajectory data set is calculated to obtain a second decision grade for each second candidate trajectory, and a second decision grade sequence is formed; wherein the first type of scene includes a plurality of ordinary driving scenes, and the second type of scene includes scenes with a target vehicle cutting in and a target vehicle cutting out.

7. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method according to any one of claims 1 to 4 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 4.

Images