Driving style recognition method, assisted driving method and device

Abstract

The present application relates to the field of automobile driving technology, and can be applied to autonomous driving vehicles, and specifically to a driving style recognition method and device, the method comprising: obtaining each feature vector value of a target vehicle at each continuous moment, the feature vector being used to represent the instantaneous driving state of the target vehicle, wherein the feature vector comprises a state feature for representing the state between the target vehicle and its related objects; sampling each feature vector value with N continuous values ​​as a sampling unit to obtain multiple samples, wherein N is a preset value; determining the driving style of the target vehicle according to the multiple samples. Based on the technical solution provided by the present application, the driving style recognition result can be made more comprehensive and accurate.

Description

Technical Field

[0001] The present application relates to the field of automobile driving technology, and in particular to a driving style recognition method, an assisted driving method, an apparatus, a computing device, and a storage medium. Background Art

[0002] On the road, different drivers may have different driving styles, such as aggressive, stable, cautious, etc. Drivers may also be in some special dangerous driving conditions, such as drunk driving, drug driving, fatigue driving, etc. Driving style is closely related to dangerous driving behavior, so accurately identifying the driving style of surrounding vehicles is very important for the safe driving of the vehicle.

[0003] Regarding the problem of driving style recognition in the prior art, the current problem is mainly the recognition of the driving style of the vehicle. The index value is calculated by collecting the driving state characteristics of the vehicle, and then the index value is compared with multiple thresholds to correspond to different driving styles. The features collected by this solution are the driving state characteristics of a single moment, which cannot well describe the continuous vehicle driving process, thus resulting in inaccurate recognition of the vehicle's driving style. Summary of the invention

[0004] In view of the above problems in the prior art, the present application provides a driving style recognition method, an assisted driving method, an apparatus, a computing device and a storage medium, which can identify the driving style of a target vehicle and identify it based on the continuity between continuous driving operations, thereby improving the accuracy of driving style recognition.

[0005] In order to achieve the above-mentioned object, the first aspect of the present application provides a driving style recognition method, comprising:

[0006] Acquire each feature vector value of the target vehicle at each consecutive moment, wherein the feature vector is used to represent the instantaneous driving state of the target vehicle;

[0007] For each of the characteristic vector values, sampling is performed with N consecutive values ​​as a sampling unit to obtain a plurality of samples;

[0008] A driving style of the target vehicle is determined based on the plurality of samples.

[0009] As described above, the technical solution provided by the present invention obtains the characteristic vector values ​​at each continuous moment and samples the characteristic vectors with time continuity, so that the obtained samples retain the continuity between continuous micro-driving operations, which can more accurately describe the driving process and thus improve the recognition accuracy of driving style.

[0010] As an implementation manner of the first aspect, when N is greater than 1, in two samples obtained by two adjacent samplings, at least one feature vector value corresponding to the same time exists in the two samples.

[0011] From the above, by making at least one feature vector value corresponding to the same moment exist in the two samples, that is, using the feature vector of the last moment in the sample obtained by the previous sampling as the feature vector of the first moment in the sample obtained by the next sampling, the correlation between the driving states in adjacent times can be more completely retained.

[0012] As an implementation manner of the first aspect, the feature vector includes a state feature representing the state of the target vehicle itself and / or a state feature representing the state between the target vehicle and its related objects.

[0013] From the above, the feature vector provided by this aspect includes not only the state features representing the state of the target vehicle itself, but also the state features representing the state between the target vehicle and its related objects, which makes the driving state description more complete and better reflects the driving state of the target vehicle, thereby making the driving style prediction more accurate.

[0014] As an implementation of the first aspect, the state characteristics representing the target vehicle's own state include at least one of the following: a state characteristic corresponding to an acceleration state, a state characteristic corresponding to a steering state, and a state characteristic corresponding to an overspeeding state.

[0015] From the above, each of the above state characteristics can be directly acquired by external vehicle sensors or cameras, making data collection simpler. In addition, since different driving styles have obvious differences in acceleration, such as aggressive drivers will frequently accelerate and decelerate, while stable drivers tend to drive at a constant speed as much as possible. Therefore, the acceleration state is an important factor in driving style identification. Secondly, different driving styles have obvious differences in steering states. For example, aggressive drivers will frequently change lanes and overtake. In addition, fatigue driving, drunk driving, drug driving, etc. will also unconsciously deviate from the lane due to loss of control ability. Therefore, the steering state is also an important factor in driving style identification. In addition, speeding can reflect the driving state of the vehicle itself. Even if there are no vehicles around, speeding is still dangerous. Therefore, this aspect uses the speeding state as an important factor in driving style identification.

[0016] As an implementation method of the first aspect, the state characteristics representing the state between the target vehicle and its related object include at least one of the following: state characteristics of the time of collision between the target vehicle and its related object, state characteristics of the relative distance between the target vehicle and its related object, and state characteristics of the relative speed between the target vehicle and its related object.

[0017] From the above, this aspect provides several possible state characteristics representing the state between the target vehicle and its related objects to describe the mutual influence between the target vehicle and the surrounding objects, which can more accurately describe the driving state of the vehicle and make driving style recognition more accurate.

[0018] As an implementation manner of the first aspect, the related object includes at least one of the following: a vehicle related to the target vehicle, a pedestrian related to the target vehicle, and a static obstacle related to the target vehicle.

[0019] As an implementation of the first aspect, the vehicle related to the target vehicle includes:

[0020] The vehicle ahead and / or behind the target vehicle in its lane when the vehicle is travelling straight ahead; or

[0021] The vehicle ahead and / or behind the target vehicle in the target lane when the vehicle changes lanes.

[0022] Based on the above, a method for determining vehicles related to a target vehicle is provided, which can accurately determine vehicles related to the target vehicle and provide support for driving style recognition.

[0023] As an implementation of the first aspect, determining the driving style of the target vehicle according to the multiple samples includes:

[0024] The driving style of the target vehicle is determined based on the multiple samples using a pre-trained latent Dirichlet distribution model.

[0025] As an implementation of the first aspect, the training process of the pre-trained latent Dirichlet distribution model includes:

[0026] Pre-acquire the characteristic vector values ​​of a preset number of target vehicles at consecutive moments;

[0027] For each of the characteristic vector values, sampling is performed with N consecutive values ​​as a sampling unit to obtain a plurality of samples;

[0028] The multiple samples are used as input layer training samples of a hidden Dirichlet distribution model, and the hidden Dirichlet distribution model is trained using a Gibbs sampling algorithm to obtain the pre-trained hidden Dirichlet distribution model.

[0029] A second aspect of the present application provides a driving style recognition device, comprising:

[0030] An acquisition module, used to acquire each feature vector value of the target vehicle at each continuous moment, wherein the feature vector is used to represent the instantaneous driving state of the target vehicle;

[0031] A sampling module, used for sampling each of the feature vector values ​​with N consecutive values ​​as a sampling unit to obtain multiple samples;

[0032] A determination module is used to determine the driving style of the target vehicle according to the multiple samples.

[0033] As an implementation manner of the second aspect, in the sampling module, when N is greater than 1, in two samples obtained by two adjacent samplings, at least one feature vector value corresponding to the same time exists in the two samples.

[0034] As an implementation manner of the second aspect, in the acquisition module and the sampling module, the feature vector includes a state feature representing the state of the target vehicle itself and / or a state feature representing the state between the target vehicle and its related objects.

[0035] As an implementation of the second aspect, the state characteristics representing the target vehicle's own state include at least one of the following: a state characteristic corresponding to an acceleration state, a state characteristic corresponding to a steering state, and a state characteristic corresponding to an overspeeding state.

[0036] As an implementation method of the second aspect, the state characteristics representing the state between the target vehicle and its related object include at least one of the following: state characteristics of the time of collision between the target vehicle and its related object, state characteristics of the relative distance between the target vehicle and its related object, and state characteristics of the relative speed between the target vehicle and its related object.

[0037] As an implementation manner of the second aspect, the related object includes at least one of the following: a vehicle related to the target vehicle, a pedestrian related to the target vehicle, and a static obstacle related to the target vehicle.

[0038] As an implementation of the second aspect, the vehicle related to the target vehicle includes:

[0039] When the vehicle is travelling straight ahead, the vehicle ahead and / or behind the target vehicle in its lane; or

[0040] When a vehicle changes lanes, the vehicle in front and / or behind is in the target lane of the target vehicle.

[0041] A third aspect of the present application provides an assisted driving method, comprising:

[0042] Determine the target vehicle relative to the vehicle according to the driving state of the vehicle;

[0043] Identify the driving style of the target vehicle using a driving style identification method provided by the first aspect;

[0044] The vehicle's driving is assisted based on the identified target vehicle's driving style.

[0045] As an implementation of the third aspect, determining a target vehicle relative to the vehicle according to the driving state of the vehicle includes:

[0046] When the vehicle is moving straight, the vehicle in front and / or behind the vehicle in the lane where the vehicle is located is the target vehicle relative to the vehicle;

[0047] When a vehicle changes lanes, the vehicle in front and / or behind located in the target lane of the vehicle is the target vehicle relative to the vehicle.

[0048] A fourth aspect of the present application provides a driving assistance device, comprising:

[0049] A target vehicle determination module is used to determine a target vehicle relative to the vehicle according to the driving state of the vehicle;

[0050] A driving style recognition device, which uses the driving style recognition method according to any one of the first aspects of the claims to recognize the driving style of the target vehicle;

[0051] The driving assistance module is used to assist the driving of the vehicle according to the driving style of the identified target vehicle.

[0052] As an implementation of the fourth aspect, determining a target vehicle relative to the vehicle according to the driving state of the vehicle includes:

[0053] When the vehicle is moving straight, the vehicle in front and / or behind the vehicle in the lane where the vehicle is located is the target vehicle relative to the vehicle;

[0054] When the vehicle changes lanes, the vehicle in front and / or behind located in the target lane of the vehicle is the target vehicle relative to the vehicle.

[0055] A fifth aspect of the present application provides a computing device, comprising:

[0056] bus;

[0057] A communication interface connected to the bus;

[0058] at least one processor connected to the bus; and

[0059] At least one memory is connected to the bus and stores program instructions, and when the program instructions are executed by the at least one processor, the at least one processor performs the instructions of the method described in any one of the first aspect or the third aspect.

[0060] The sixth aspect of the present application provides a computer-readable storage medium having program instructions stored thereon, wherein when the program instructions are executed by a computer, the computer executes instructions of the method described in any one of the first aspect or the third aspect.

[0061] The technical solution provided in the present application obtains the feature vector values ​​at each continuous moment and samples the feature vectors with time continuity, so that the obtained samples retain the continuity between consecutive onlooker driving operations, which can more accurately and fully describe the driving process, provide richer information for driving style recognition, and thus improve the recognition accuracy of driving style.

[0062] These and other aspects of the present application will become more apparent from the following description of the embodiment(s). BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The following further illustrates the various features of the present application and the connections between the various features with reference to the accompanying drawings. The accompanying drawings are all exemplary, and some features are not shown in actual proportion. Some drawings may omit features that are conventional in the field involved in the present application and are not necessary for the present application, or additional features that are not necessary for the present application may be shown. The combination of the various features shown in the accompanying drawings is not intended to limit the present application. In addition, throughout this specification, the same figure numerals refer to the same content. The specific description of the drawings is as follows:

[0064] Figure 1 A schematic diagram of target vehicle determination provided in an embodiment of the present application;

[0065] Figure 2 A schematic diagram of obtaining characteristic parameters of a driving style classification scheme based on a supervised machine learning method in the second prior art;

[0066] Figure 3 A flowchart of a neural network training scheme for classifying driving styles based on a supervised machine learning method in the second prior art;

[0067] Figure 4 This is a driving style classification result diagram obtained based on the K-means clustering algorithm in the prior art three;

[0068] Figure 5 A schematic diagram of the structure of an autonomous driving vehicle provided in an embodiment of the present application;

[0069] Figure 6 A flowchart of a driving style identification method provided in an embodiment of the present application;

[0070] Figure 7 A flowchart of a driving style identification method provided by another embodiment of the present application;

[0071] Figure 8 a, b, and c are schematic diagrams of sampling results of serialized sampling when N=1, N=2, and N=3, respectively, provided in the embodiments of the present application;

[0072] Fig. 9a and b are schematic diagrams of sampling results at different sampling frequencies provided in the embodiments of the present application;

[0073] Fig.10 A serialized sampling effect diagram provided in an embodiment of the present application;

[0074] Fig.11 A probability graph of an implicit Dirichlet distribution model provided in an embodiment of the present application;

[0075] Fig.12 A driving style probability distribution diagram provided in an embodiment of the present application;

[0076] Fig.13 A schematic diagram of the structure of a driving style recognition device provided in an embodiment of the present application;

[0077] Fig.14 A schematic diagram of a specific implementation method of a driving style recognition device provided in an embodiment of the present application;

[0078] Fig.15 A flowchart of an assisted driving method provided in an embodiment of the present application;

[0079] Fig.16 A schematic diagram of the structure of a driving assistance device provided in an embodiment of the present application;

[0080] Fig.17 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0081] The words "first, second, third" etc. or similar terms such as module A, module B, module C, etc. in the specification and claims are only used to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that the specific order or sequence can be interchanged where permitted so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0082] In the following description, the numbers representing the steps, such as S110, S120, etc., do not necessarily mean that the steps must be executed in this manner. If permitted, the order of the previous and next steps can be interchanged, or the steps can be executed simultaneously.

[0083] The term "comprising" as used in the description and claims should not be interpreted as being limited to what is listed thereafter; it does not exclude other elements or steps. Therefore, it should be interpreted as specifying the presence of the features, integers, steps or components mentioned, but does not exclude the presence or addition of one or more other features, integers, steps or components and groups thereof. Therefore, the expression "a device comprising means A and B" should not be limited to a device consisting of components A and B only.

[0084] References to "one embodiment" or "an embodiment" in this specification mean that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present application. Therefore, the phrases "in one embodiment" or "in an embodiment" appearing in various places in this specification do not necessarily refer to the same embodiment, but may refer to the same embodiment. In addition, in one or more embodiments, the particular features, structures, or characteristics can be combined in any appropriate manner, as would be apparent to one of ordinary skill in the art from this disclosure.

[0085] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present application. In case of any inconsistency, the meaning described in this specification or the meaning derived from the contents recorded in this specification shall prevail. In addition, the terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0086] Before further describing the specific implementation methods of the present application in detail, the nouns and terms involved in the embodiments of the present application, as well as their corresponding uses, effects, functions, etc. in the present application are described. The nouns and terms involved in the embodiments of the present application are subject to the following interpretations:

[0087] Latent Dirichlet Allocation (LDA): Latent Dirichlet Allocation is a commonly used topic model that can give the topic of each document in a document set in the form of probability distribution.

[0088] Topic model: Topic model is mainly used in natural language processing to identify the topic of a document in a series of documents. Latent Dirichlet allocation is a common topic model. In the embodiment of the present application, the mathematical form of the topic model is used to identify the driving style of the vehicle from various driving characteristics of the vehicle over a period of time.

[0089] In order to accurately describe the technical content of the application and to accurately understand the application, the following explanations or definitions are given for the terms used in this specification before describing the specific implementation methods.

[0090] Time-To-Collision (TTC): Time-To-Collision is a widely used indicator in risk estimation when vehicles are driving. It is defined as the quotient of the relative distance between the target vehicle and the vehicles around the target vehicle divided by the relative speed.

[0091] Autonomous Vehicle (AV): Autonomous Vehicle mainly relies on the coordinated cooperation of artificial intelligence, visual computing, radar, monitoring devices and global positioning system to realize unmanned driving of intelligent cars through computer systems.

[0092] Driving style: refers to the way of driving or the habitual driving method, including the choice of driving speed, the choice of driving distance, etc.

[0093] Driving characteristics: variables used to describe the state of the vehicle at a certain moment, represented by feature vectors.

[0094] Sequence sample: The sequence sample in the embodiment of the present application is a sequence segment composed of driving features at multiple consecutive moments, and is the smallest unit in the sample set input into the topic model.

[0095] Target vehicle: For this vehicle, the target vehicle is the vehicle that can affect the driving of this vehicle. Figure 1 As shown, when the vehicle's route is straight, the target vehicle at this time is the vehicle in front of the same lane. When the vehicle is about to change lanes, there are two target vehicles at this time, namely the front and rear vehicles of the target lane. If the target lane vehicle is parallel to the vehicle, it can be regarded as the front or rear vehicle of the target lane. It should be noted that only one method of screening target vehicles is given here. In other embodiments, there may be other methods of screening target vehicles. For example, when the vehicle is going straight, the rear vehicle in the same lane can also be taken as one of the target vehicles. For example, when the vehicle is about to change lanes, the front and rear vehicles in the current lane can also be taken as one of the target vehicles at the same time.

[0096] N-order serial sampling: In the embodiment of the present application, it means that among a plurality of continuous driving state feature vectors s, N continuous driving state feature vectors s are sampled as a group, and each group of sequence segments is used as a data sample, and there is at least one crossed driving state feature vector s between adjacent data samples. For example, the following formula is an example of a crossed driving state feature vector between adjacent data samples w1 and w2:

[0097]

[0098] You can also refer to Figure 8 The embodiments shown in (b) and (c) are examples in which adjacent data samples have a crossed driving state feature vector, that is, in adjacent data samples, the last feature vector in the previous data sample is also used as the first feature vector in the next data sample. The data samples constructed in this way can reflect the correlation between the driving states at corresponding adjacent moments.

[0099] Below, we first analyze the existing technology:

[0100] Prior art 1 provides a method based on driving style recognition (Murphey YL, Milton R, Kiliaris L. Driver's style classification using jerk analysis [C] / / 2009 IEEE Workshop on Computational Intelligence in Vehicles and Vehicular Systems. IEEE, 2009: 23-28.), which is a classification method based on rule description. The method first uses the acceleration change rate as a measure of driving style, and divides the driving style into four types: aggressive, normal, calm, and parking by setting a threshold range for the acceleration change rate; then, the acceleration change rate of the vehicle is collected, and it is determined which range of the acceleration change rate ranges corresponding to the above four types the collected acceleration change rate falls into, so as to obtain the driving style of the vehicle.

[0101] First, this scheme is to identify the driving style of the vehicle by collecting the acceleration change rate of the vehicle. In addition, the scheme is relatively simple in its way of dividing driving styles, which only abstracts driving style identification as a simple classification problem and cannot describe the possibility of different driving styles of vehicles. Secondly, this scheme only distinguishes driving styles by setting thresholds, and does not provide a method for setting thresholds, nor does it provide a verification rule setting for correctness. In addition, this method only considers the acceleration change rate as a factor in judging driving style, which has great limitations and cannot fully reflect the driving status of the vehicle.

[0102] Prior art 2 provides a method based on supervised machine learning to classify driving styles (CN108629372A), such as Figure 2 The flowchart of obtaining the driving style characteristic parameters in the scheme is shown, which mainly includes the following steps:

[0103] a: The driver selects the type of simulated driving condition through the interface of the simulated driving interaction module 14;

[0104] b: The main control module 13 returns confirmation information to the interface of the simulated driving interaction module 14, and provides the preset driving state under the selected driving condition and the initial driving state of the operating vehicle;

[0105] c: The driver operates the accelerator pedal 12 and the brake pedal 11 to control the operation of the vehicle and collect characteristic parameters during the operation.

[0106] d: The collected characteristic parameters are transmitted to the main control module 13, and the main control module 13 sends the characteristic vector to the host computer 3, and the host computer 3 uses the neural network algorithm to identify the driving style.

[0107] In addition, the method also provides a flowchart for identifying driving style using BPNN (neural network) algorithm, which includes Figure 3 The training process of the BPNN algorithm model shown in FIG. 1 mainly includes the following steps:

[0108] S1: Using the experimental system to collect characteristic parameters of multiple sets of set periods of driving testers with different styles and types under different simulated driving conditions as training data. At the same time, using the experimental system to collect characteristic parameters of multiple sets of set periods of driving testers under different simulated driving conditions as test data;

[0109] S2: Normalize the training data;

[0110] S3: Establish a BPNN algorithm model with a tutor type and set the training parameters of the BPNN algorithm model;

[0111] S4: Use the training data to train the BPNN algorithm model until the training target error is within the set range.

[0112] First, this scheme is to classify the driving style of the vehicle by collecting the characteristic parameters of the vehicle when it is running. Secondly, this scheme adopts a supervised machine learning method, which requires the training data set to be labeled. Among them, the labeling of driving styles in the training data set is subjective and requires a lot of manpower; in addition, the characteristic parameters selected in this scheme are accelerator pedals, brake pedals, etc. These characteristic parameters require the actual operation of the driver as input, and it is difficult to obtain the characteristic parameters of the vehicle through external sensors such as sensors of other vehicles or sensors of roadside equipment. Therefore, it can only be used to judge the driving style of the vehicle, and is not suitable for judging the driving style of surrounding vehicles on the road.

[0113] The third prior art provides a method based on unsupervised machine learning to classify driving styles (Li Jingwei, Zhao Zhiguo, Shen Peihong, et al. Research on K-means clustering and identification method of driving style [J]. Automobile Technology, 2018 (12): 8-12.). This method is based on clustering algorithms such as K-means to classify driving behaviors with similar characteristics into one driving style. Among them, the characteristics are acceleration, acceleration average, and acceleration standard deviation. Figure 4 The figure shows the driving style classification results obtained by this method based on the K-means clustering algorithm, where the principal components of each coordinate axis represent different feature parameters considered by the model, and each cluster represents a type of driving style.

[0114] First, this solution is the same as the above-mentioned prior art, which is a solution to predict the driving style of the vehicle by collecting the characteristic parameters of the vehicle. Secondly, this solution uses an unsupervised machine learning method combined with a clustering algorithm, which is essentially a classification problem and cannot describe the possibility of different driving styles of vehicles. In addition, the features in this solution are all motion features, such as acceleration, acceleration average, and acceleration standard deviation, which neither considers the interaction between the vehicle and the surrounding environment, nor the correlation between the driving states of the vehicle in adjacent time. However, when the vehicle is driving, the surrounding environment and surrounding vehicles have a huge impact on its safe driving.

[0115] Based on the research on the prior art and the defects of the prior art, an embodiment of the present application provides a driving style recognition method, which can make the recognition more accurate by collecting feature vectors with time continuity to identify the driving style of the target vehicle.

[0116] The embodiments of the present application are described in detail below in conjunction with the accompanying drawings. First, the application scenario of the driving style recognition method provided by the embodiments of the present application is introduced.

[0117] The driving style recognition method provided in the embodiment of the present application is applicable to autonomous vehicles (AV, Autonomous Vehicle). The application scenario is that when the autonomous vehicle is driving on the road, it can make real-time judgments on the driving styles of other vehicles on the road, predict the trajectories of other vehicles at the next moment based on the judgment results, and provide auxiliary information for the decision-making planning of the vehicle.

[0118] like Figure 5 As shown, it is a schematic diagram of the structure of an autonomous driving car used in the driving style recognition solution provided in an embodiment of the present application, which includes a vehicle-mounted sensor for sensing the external environment, a data calculation unit, a data storage unit CPU and / or a GPU and other computing units. Among them, the vehicle-mounted sensor for sensing the external environment includes but is not limited to a camera, a laser radar (LiDAR), an ultrasonic radar, a millimeter-wave radar, etc. The data calculation unit includes but is not limited to a mobile data center (MDC, Mobile Data Center, MDC). In one embodiment of the present application, the vehicle-mounted sensor is used to collect driving state information of other vehicles on the road. The data calculation unit is used to process the collected driving state information, discretize it, serialize and sample it, and perform driving style recognition based on the LDA model. The data storage unit is used to store the collected driving state data and LDA model related parameters. The LDA model is trained online or offline using a CPU or GPU, etc.

[0119] Referring to the figures below, a driving style recognition method provided in an embodiment of the present application is described in detail.

[0120] See Figure 6 The flowchart shown in the figure illustrates an embodiment of a driving style recognition method provided by the present application. The method mainly includes steps S110-S130. Each step is introduced in turn below:

[0121] S110: Acquire each feature vector value of the target vehicle at each consecutive moment, wherein the feature vector is used to represent the instantaneous driving state of the target vehicle.

[0122] In this step, the feature vector representing the instantaneous driving state of the target vehicle may include a state feature representing the target vehicle's own state, and / or a state feature representing the state between the target vehicle and its related objects. It can be understood that the state features constitute the feature vector.

[0123] The state characteristics representing the state of the target vehicle itself may include but are not limited to at least one of the following: state characteristics corresponding to the acceleration state, state characteristics corresponding to the steering state, and state characteristics corresponding to the speeding state. The state characteristics representing the state between the target vehicle and its related objects may include but are not limited to at least one of the following: state characteristics of the time of collision between the target vehicle and its related objects, state characteristics of the relative distance between the target vehicle and its related objects, and state characteristics of the relative speed between the target vehicle and its related objects.

[0124] In addition, the related objects may include but are not limited to at least one of the following: a vehicle related to the target vehicle, a pedestrian related to the target vehicle, and a static obstacle related to the target vehicle.

[0125] In this embodiment, the vehicles related to the target vehicle include:

[0126] The vehicle ahead and / or behind the target vehicle in its lane when the vehicle is travelling straight ahead; or

[0127] The vehicle ahead and / or behind the target vehicle in the target lane when the vehicle changes lanes.

[0128] S120: Sampling each of the feature vector values ​​by taking N consecutive values ​​as a sampling unit to obtain a plurality of samples.

[0129] In this step, when N is equal to 1, each sample obtained by sampling contains a feature vector value. Optionally, when N is greater than 1, there is no feature vector value corresponding to the same moment in the two samples obtained by two adjacent samplings, that is, the samples obtained by sampling sequentially at the sampling moment do not have cross-overlapping feature vectors. Optionally, when N is greater than 1, there is at least one feature vector value corresponding to the same moment in the two samples obtained by two adjacent samplings, that is, there is at least one cross-overlapping feature vector in the samples obtained by two adjacent samplings.

[0130] S130: Determine the driving style of the target vehicle according to the multiple samples.

[0131] The plurality of samples are input into a pre-trained driving style prediction model to obtain the output of the model, which is the driving style of the target vehicle. The output is described in the form of a multi-dimensional vector to represent the possible probabilities of various driving styles. Optionally, the driving style with the highest probability is taken as the prediction result.

[0132] In addition, the pre-trained driving style prediction model can be predicted using a topic model, such as an LDA model, a pinball machine distribution model, etc.; or a deep neural network can be used.

[0133] Referring to the figures below, a driving style recognition method provided in another embodiment of the present application is described in detail.

[0134] See Figure 7 The flowchart shown in the figure illustrates an embodiment of a driving style recognition method provided by the present application. The method mainly includes steps S210-S230, and each step is introduced in turn below:

[0135] S210: Constructing a feature vector for representing the instantaneous driving state of the target vehicle; wherein the instantaneous driving state of the target vehicle includes the state between the target vehicle and surrounding related objects.

[0136] In this step, the various driving state features after the above discretization process are taken as a feature vector s, which can be expressed as s = (acc, dir, ttc, overspd), and is used to describe the instantaneous driving state of the target vehicle.

[0137] Among them, the driving state of the target vehicle can be represented by a multi-dimensional feature vector. In this embodiment, considering the four dimensions of the target vehicle's acceleration state, steering state, related vehicle TTC, and speeding state, a four-dimensional feature vector s is used to represent the instantaneous driving state of the target vehicle. The feature vector s can be expressed as s = (acc, dir, ttc, overspd), wherein the definition of each vector element of the feature vector and its discretization value method will be described in detail below.

[0138] In this embodiment, in the selection of driving state features, driving state features that can represent the driving style of the vehicle are selected, including but not limited to: the acceleration state acc (acceleration) of the target vehicle, the steering state dir (direction) of the target vehicle, and the overspeed state overspd (overspeed) of the target vehicle; in this embodiment, in the selection of driving state features, the state between the vehicle and the surrounding vehicles is also considered. Therefore, the driving state features of this embodiment also include the collision time TTC (Time-To-Collision) between the target vehicle and the surrounding vehicles of the target vehicle. In addition, other factors that affect the state between the vehicle and the surrounding vehicles may also be included, for example, different lanes have different speed limits on certain lanes (such as highway lanes), and may also include the direction of whether the target vehicle is in the left lane or the right lane. This embodiment only takes the collision time TTC between the target vehicle and the surrounding vehicles of the target vehicle as an example, and its purpose is to reduce redundant state features in the LDA model. In addition, the driving state features may also consider the collision time between the target vehicle and pedestrians or the collision time between the target vehicle and static obstacles on the road.

[0139] In some embodiments, different driving styles have obvious differences in acceleration performance. For example, aggressive drivers will frequently accelerate and decelerate, while stable drivers tend to drive at a constant speed as much as possible. Therefore, the acceleration state is an important factor in driving style identification. In this embodiment, the acceleration state of the target vehicle is represented by the state feature acc. The acceleration state of the target vehicle is discretized according to the acceleration state of the target vehicle. For example: when the acceleration of the target vehicle is greater than 0, that is, the target vehicle is in an accelerating state, its discrete value is 1; when the acceleration of the target vehicle is equal to 0, that is, the target vehicle is in a uniform speed state, its discrete value is 0; when the acceleration of the target vehicle is less than 0, that is, when the target vehicle is in a decelerating state, its discrete value is -1. See the following formula for details:

[0140]

[0141] Among them, a is the acceleration value of the target vehicle collected in real time.

[0142] Another example: the acceleration state of the target vehicle can be more finely discretized by setting thresholds. For example: when the acceleration of the target vehicle is greater than the first acceleration threshold a1, that is, the target vehicle is in the first acceleration state, its discrete value is 2; when the acceleration of the target vehicle is not greater than the first acceleration threshold a1 and is greater than 0, that is, the target vehicle is in the second acceleration state, its discrete value is 1; when the acceleration of the target vehicle is equal to 0, that is, the target vehicle is in a uniform speed state, its discrete value is 0; when the acceleration of the target vehicle is less than 0 and not less than the negative first acceleration threshold -a1, that is, the target vehicle is in the first deceleration state, its discrete value is -1; when the acceleration of the target vehicle is less than the negative first acceleration threshold -a1, that is, the target vehicle is in the second deceleration state, its discrete value is -2. See the following formula for details:

[0143]

[0144] Wherein, a is the acceleration value of the target vehicle collected in real time, and a1 is the first acceleration threshold, which is a preset value.

[0145] The above provides two methods for discretizing the acceleration state of the target vehicle. In practical applications, if a more refined discretization is required, it can be achieved by setting multiple thresholds, which will not be described in detail in this article.

[0146] In some embodiments, different driving styles have obvious differences in steering states. For example, aggressive drivers will frequently change lanes to overtake. In addition, fatigue driving, drunk driving, drug driving, etc. will also unconsciously deviate from the lane due to loss of control ability. Therefore, the steering state is also an important factor in driving style identification. Therefore, the steering state of the target vehicle is selected as one of the driving state features, represented by dir.

[0147] The process of discretizing the steering state of the target vehicle can be:

[0148] When the target vehicle is turning left, its discrete value is 1; when the target vehicle is going straight, its discrete value is 0; when the target vehicle is turning right, its discrete value is -1. For details, please refer to the following formula:

[0149]

[0150] In some embodiments, the acceleration state and the steering state cannot fully reflect the impact of the driving action on the surrounding vehicles. Another prominent feature of aggressive drivers is that when the collision time TTC between the target vehicle and the vehicles around the target vehicle is small, the target vehicle continues to perform maneuvering operations. Therefore, the state variable ttc of the collision time between the target vehicle and the vehicles around the target vehicle is selected as one of the driving state characteristics.

[0151] In this embodiment, the collision time TTC between the target vehicle and the surrounding vehicles of the target vehicle is defined as the quotient of the relative distance between the target vehicle and the surrounding vehicles of the target vehicle and the relative speed of the two. When the quotient value is smaller, it means that the time available for the driver to react is shorter, which is generally considered to be more dangerous. In addition, in other embodiments, the collision time TTC between the target vehicle and the surrounding vehicles of the target vehicle can also only consider the relative distance between the two or only consider the relative speed between the two. In this implementation, the quotient of the two is taken as an example, and specifically refer to the following formula:

[0152]

[0153] Among them, Δx is the relative distance between the target vehicle and the vehicles around the target vehicle, that is, the distance between the two vehicles, and Δv is the relative speed between the target vehicle and the vehicles around the target vehicle.

[0154] When the vehicle is in a straight-moving state, the TTC of the target vehicle is the TTC between the target vehicle and the vehicle in front of it in the same lane; when the vehicle is in a lane-changing state, it may involve the TTC between the target vehicle and the vehicle in front of it and the vehicle in front of it in the lane mentioned above. At this time, the minimum value of the two is taken as the judgment indicator. See the following formula for details:

[0155] TTC≈min(TTC front ,TTC rear )

[0156] Among them, TTC front is the TTC between the target vehicle and the vehicle ahead in the target lane, TTC rear It is the TTC between the target vehicle and the vehicle behind it in the target lane.

[0157] The process of discretizing the collision time TTC between the target vehicle and the vehicles around the target vehicle can be:

[0158] When the collision time TTC between the target vehicle and the surrounding vehicles of the target vehicle is greater than the first preset threshold, its discrete value is 0; when the collision time TTC between the target vehicle and the surrounding vehicles of the target vehicle is not greater than the first preset threshold, its discrete value is 1. See the following formula for details:

[0159]

[0160] Among them, TTC th is the first preset threshold, and ttc is the state characteristic of the collision occurrence time between the target vehicle and the vehicles around the target vehicle.

[0161] In addition, in other embodiments, the refinement of the feature discretization process may be improved by setting multiple thresholds.

[0162] In some embodiments, the speeding status may reflect the driving status of the vehicle itself. Even if there are no other vehicles around, speeding is still dangerous. Therefore, the speeding status is an important basis for identifying the driving style.

[0163] The process of discretizing the overspeed state can be:

[0164] When the speed of the target vehicle is greater than the second preset threshold, its discrete value is 0; when the speed of the target vehicle is not greater than the second preset threshold, its discrete value is 1. For details, see the following formula:

[0165]

[0166] Among them, v is the speed value of the target vehicle collected in real time, v th is the second preset threshold, and overspd is the state characteristic of the overspeed state.

[0167] In addition, in other embodiments, the refinement of the feature discretization process may be improved by setting multiple thresholds.

[0168] Based on the results of the discrete processing of each driving state provided by the above embodiment, when the state characteristic acc of the acceleration state of the target vehicle and the state characteristic dir of the steering state of the target vehicle respectively use three discrete values, and the state characteristic ttc of the time of collision between the target vehicle and the vehicles around the target vehicle and the state characteristic overspd of the speeding state of the target vehicle respectively use two discrete values, the driving state has a total of 3*3*2*2=36 possible values. It should be noted here that the above embodiment only provides one or two possible implementation methods of discretization, which can be adjusted according to actual conditions. The adjustment here can be either the selection of driving state characteristics or the degree of discreteness.

[0169] S220: Sampling the target vehicle, with each N consecutive samplings being regarded as a sequence sample, i.e., performing serial sampling.

[0170] Specifically: within a preset time length T, that is, within the observation time window length T, the feature vector s of the target vehicle is serially sampled with a preset sampling frequency f and a preset order N to obtain multiple data samples of the target vehicle; the multiple data samples are used as a data sample set, and the sample set is used to describe the driving state of the target vehicle in the past period of time.

[0171] In this embodiment, the serialized sampling is an N-order serialized sampling, that is, N consecutive feature vectors s constitute a sequence sample w (ie, a data sample), which can be expressed as w=(s1, s2, ...s N ) TWherein, N represents the order of serialized sampling, and the size of N can be selected according to the actual situation. The larger N is, the more detailed the description of the driving process is. However, the larger N is, the greater the amount of calculation is, and the time cost of the early training model is also higher. For example, in the above S110, if there are 36 possible values ​​of the driving state, correspondingly, there are 36 samples obtained by N-order sampling. N There are possibilities, where T represents the preset time length, i.e., the time length for collecting the driving state of the target vehicle. Optionally, in the process of serialized sampling, N feature vectors are taken as a data sample. When N is 1, in the data sample set, one feature vector is taken as a data sample, i.e., w1=(s1), w2=(s2). When N is an integer greater than 1, in the data sample set, the feature vectors are sampled in sequence, and N feature vectors are taken as a data sample. For example: when N=2, the data samples under this sampling method are: w1=(s1, s2), w2=(s3, s4). When N is an integer greater than 1, in the data sample set, at least one feature vector at the tail of a data sample can also be taken as the head feature vector in the next adjacent data sample. For example: when N=2, the data samples under this sampling method are: w1=(s1, s2), w2=(s2, s3).

[0172] like Figure 8 a, b, and c are the sampling results when N = 1, N = 2, and N = 3, respectively. In the figure, each car represents a driving state feature vector at a moment, that is, an s, and each circle represents a sequence sample w composed of N driving state feature vectors.

[0173] Figure 8 The first-order serialized sampling result (i.e., data sample set) corresponding to a is (w1, w2, w3, w4, w5); among which w1 = (s1), w2 = (s2), w3 = (s3), w4 = (s4), w5 = (s5).

[0174] Figure 8 The second-order serialized sampling result corresponding to b is (w1, w2, w3, w4); where w1 = (s1, s2), w2 = (s2, s3), w3 = (s3, s4), w4 = (s4, s5).

[0175] Figure 8 The third-order serialized sampling result corresponding to c is (w1, w2), where w1 = (s1, s2, s3) and w2 = (s3, s4, s5).

[0176] In this embodiment, another optional variable in the N-order serial sampling is the sampling frequency f. Fig. 9As shown in a and b, each car in the figure represents a driving state feature vector at a moment, that is, an s, and each circle represents a sequence sample w composed of N driving state feature vectors. Fig. 9 The sampling frequency of a is f1, Figure 8 b uses a higher sampling frequency f2=2f1, Fig. 9 The sampling result of a is: w1 = (s1, s2, s3). Fig. 9 The sampling results of b are w1 = (s1, s2, s3), w2 = (s3, s4, s5). It can be seen that the higher the sampling frequency, the more sequence samples w are obtained, the more detailed the description of the target vehicle's driving process is, and the final recognition result can be more accurate. However, the higher the sampling frequency, the larger the sequence samples w are, which will also lead to a larger amount of online calculation of the LDA model.

[0177] Based on the serialized sampling process of this step, the correlation between driving states in adjacent time periods can be preserved. Fig.10 As shown, the direction of the arrow represents the target lane direction of the vehicle, that is, the turning state. The 4 small matrices in the left figure represent the driving data at 4 moments, and the 4 small matrices in the right figure represent the driving data at 4 moments. Since the LDA model does not consider the order in which the samples appear, the sampling results shown in the figure are independent of the order in which the samples appear. Let N = 1, that is, first-order sampling, and sample the driving states of the left and right pictures. The sampling results corresponding to the left picture are a, b, c, c, and the sampling results corresponding to the right picture are a, b, c, c; let N = 2, that is, second-order sampling, and sample the driving states of the left and right pictures. The sampling results corresponding to the left picture are (a, c), (c, b), (b, c), and the sampling results corresponding to the right picture are (a, c), (c, a), (b, c); let N = 3, that is, third-order sampling, and sample the driving states of the left and right pictures. The sampling results corresponding to the left picture are (a, c, b), (c, b, c), and the sampling results corresponding to the right picture are (b, c, a), (c, a, c). It can be seen from the above sampling results that in the first-order sampling, there is no difference in the sampling results of the left and right pictures. In the second-order sampling, there is one sampling sample that is different between the two. In the third-order sampling, both sampling samples are different. It shows that when the serialization sampling order increases, the correlation between driving states will be preserved. This correlation is closely related to driving style. Therefore, the serialization sampling process provided in this step can more fully describe the driving process, provide richer information for driving style recognition, and help improve the accuracy of driving style recognition.

[0178] S230: Determine the driving style of the target vehicle based on the data sample set, ie, the driving state of the target vehicle, and the trained LDA model.

[0179] Specifically, the data sample set is used as the input of the pre-trained LDA model to obtain the probabilities of various driving styles output by the LDA model, and the driving style with the highest probability is the determined driving style of the target vehicle.

[0180] In this embodiment, the model used for driving style prediction is not limited to the LDA model, but can also be a pinball machine distribution model or a deep neural network, etc.

[0181] Wherein, the implicit Dirichlet distribution model can be a model for offline training. The driving state data of roadside vehicles can be collected by cameras or other data acquisition instruments, or a large amount of vehicle driving state data collected online by autonomous driving vehicles. The large amount of collected data is discretized and serialized in the aforementioned embodiment to obtain a large number of input samples, which are used as input samples of the implicit Dirichlet distribution model to train the model. Wherein, the training algorithm can be selected according to actual conditions. For example: Gibbs Sampling, variational inference, maximum expectation algorithm (EM, Expectation-Maximization algorithm) and other relatively mature topic model training algorithms. It should be noted here that the model is an offline trained model during use, but the model also allows real-time training and correction of online sample collection to improve the accuracy of the model recognition.

[0182] like Fig.11 As shown, it is a probability diagram of the implicit Dirichlet distribution model provided by this embodiment. In the figure, M is the total number of target vehicles, T is the preset time length (i.e., the time length for collecting the driving status of a certain target vehicle), m is the mth target vehicle, and w t，m is the t-th driving state sequence sample of the m-th target vehicle, z t，m is the driving style corresponding to the t-th driving state sequence sample of the m-th target vehicle, is the driving style feature vector of the mth target vehicle, i.e. z t，m satisfy distribution, α is The prior distribution parameter of is the preset value. is the probability distribution of driving state sequence samples under the kth driving style, that is, w t，m satisfy distributed.

[0183] in, p k is the probability of the kth driving style of the target vehicle, K is the total number of target vehicle driving styles, satisfying and q dis the probability of the dth driving state sequence sample, D is the total number of driving state sequence samples, satisfying And q d ≥0.

[0184] based on Fig.11 The probability diagram of the model shown in the figure takes the calculation process of the Gibbs sampling algorithm as an example to illustrate the process of obtaining various quantities during training of the model:

[0185] A large amount of vehicle driving status data is collected, and the data is discretized and serialized into N orders to obtain a large number of training input samples.

[0186] Randomly assign a number to each sample in the input sample. For example, assign the driving style z corresponding to the t-th driving state sequence sample of the m-th target vehicle to t，m The number is k, where k∈K, K is the total number of target vehicle driving styles.

[0187] The driving style corresponding to each sample is sampled using the Gibbs sampling formula and the sampling results are updated.

[0188] Repeat the sampling and updating process using the Gibbs sampling formula in the previous step until the Gibbs sampling algorithm reaches the convergence condition. At this time, the frequency of each target vehicle's driving state sequence sample corresponding to the driving style of each target vehicle's driving state sequence sample is counted, and it can be obtained By counting the frequency of driving styles corresponding to each driving state sequence sample of each target vehicle, we can get Due to the parameters It is related to the driving style of the vehicles in the training database and is not helpful for predicting new target vehicles. Therefore, the training process obtains Parameter information usually does not need to be saved.

[0189] In addition, α is a preset prior distribution parameter.

[0190] based on Fig.11 The probability diagram of the model described above illustrates the process of obtaining various quantities during actual prediction by the model:

[0191] α is the preset value, w t，m In order to obtain the driving state sequence samples of the target vehicle in real time, is the value obtained through offline training, z t，m is a randomly initialized value, is the output value, that is, the driving style probability corresponding to the target vehicle obtained by the output prediction.

[0192] In this embodiment, the role of the implicit Dirichlet distribution model is to analyze each driving data sample set, count the frequency of different driving states in the sample set, and determine which driving styles are included in the currently collected driving sample set and the proportion of each driving style, that is, probability, based on the statistical results. The number of driving styles K needs to be preset before identification.

[0193] Based on the implicit Dirichlet distribution model in this step, the driving style recognition results can be output in the form of feature vectors, that is, each driving style can be presented in the form of probability. Assuming that there are 6 driving styles, namely: aggressive, stable, cautious, tired, drunk, and drugged, the model is used to identify the driving style of the target vehicle, and the output result is like Fig.12 As shown in the figure, the probability of the target vehicle being aggressive is 0.3, the probability of being stable is 0.05, the probability of being cautious is 0.05, the probability of being tired is 0.1, the probability of being drunk is 0.3, and the probability of being drugged is 0.2. This recognition result presented in the form of probability is not just a simple classification of driving style, but can provide more information and data for subsequent control.

[0194] Below, refer to Fig.13 and Fig.14 To illustrate an embodiment of the present application, a driving style online identification device is provided, such as Fig.13 As shown, the device includes an acquisition module, a sampling module and a determination module.

[0195] The acquisition module is used to obtain each feature vector value of the target vehicle at each continuous moment, and the feature vector is used to represent the instantaneous driving state of the target vehicle.

[0196] Specifically, the acquisition module may also include Fig.14 The initialization module, driving state feature acquisition module and discretization processing module shown. The initialization module is used to initialize the sampling frequency f, the time length T for acquiring the driving state of the target vehicle, the order N of the serialized sampling and the model parameters of the pre-trained LDA model. The driving state feature acquisition module is used to acquire various driving state features of the target vehicle. The discretization processing module is used to discretize the driving state features and use all the driving state features after the discretization as a feature vector, that is, to obtain a feature vector for representing the instantaneous driving state of the target vehicle.

[0197] The sampling module is used to sample each of the feature vector values ​​by taking N consecutive values ​​as a sampling unit to obtain multiple samples.

[0198] Specifically, the sampling module is Fig.14The serialization sampling module shown is used to perform serialization sampling on the feature vector to obtain the data sample after the serialization sampling.

[0199] A determination module is used to determine the driving style of the target vehicle according to the multiple samples.

[0200] Specifically, the determination module is Fig.14 The driving style online recognition module shown is used to determine the driving style of the target vehicle based on data samples.

[0201] If it is determined that the driving style of the target vehicle uses the LDA model, the driving style online identification device may further include an LDA model online training module, which is used to modify the LDA model in real time online using an online variational inference algorithm.

[0202] In this embodiment, in the sampling module, when N is greater than 1, in two samples obtained by two adjacent samplings, at least one feature vector value corresponding to the same time exists in the two samples.

[0203] In the acquisition module and the sampling module, the feature vector includes a state feature representing the state of the target vehicle itself and / or a state feature representing the state between the target vehicle and its related objects.

[0204] The state feature representing the state of the target vehicle itself includes at least one of the following: a state feature corresponding to an acceleration state, a state feature corresponding to a steering state, and a state feature corresponding to an overspeeding state.

[0205] The state characteristics representing the state between the target vehicle and its related object include at least one of the following: state characteristics of the time of collision between the target vehicle and its related object, state characteristics of the relative distance between the target vehicle and its related object, and state characteristics of the relative speed between the target vehicle and its related object.

[0206] The related objects include at least one of the following: a vehicle related to the target vehicle, a pedestrian related to the target vehicle, and a static obstacle related to the target vehicle.

[0207] The vehicles related to the target vehicle include:

[0208] The vehicle in front of and / or behind the target vehicle’s lane when the vehicle is traveling straight; or the vehicle in front of and / or behind the target vehicle’s lane when the vehicle is changing lanes.

[0209] like Fig.15 As shown, an assisted driving method provided by an embodiment of the present application mainly includes steps S310-S330. Each step is introduced in sequence below: including:

[0210] S310: Determine a target vehicle relative to the vehicle according to the driving state of the vehicle.

[0211] In this step: when the vehicle is moving straight, the vehicle in front and / or behind the vehicle in the lane where the vehicle is located is the target vehicle relative to the vehicle;

[0212] When the vehicle changes lanes, the vehicle in front and / or behind located in the target lane of the vehicle is the target vehicle relative to the vehicle.

[0213] It should be noted here that the method for determining the target vehicle relative to the vehicle provided in this step is only an example, and in other embodiments, it can be set according to actual needs. For example, when the vehicle is moving straight, the vehicle parallel to the vehicle can also be used as the target vehicle relative to the vehicle. When the vehicle changes lanes, the vehicle in front and / or the vehicle behind the vehicle in the lane where the vehicle is located can also be used as the target vehicle relative to the vehicle.

[0214] S320: The driving style of the target vehicle is identified using a driving style identification method provided in the above embodiment, so this step will not be described in detail.

[0215] S330: Assisting the driving of the vehicle according to the identified driving style of the target vehicle.

[0216] like Fig.16 As shown, an auxiliary driving device provided by an embodiment of the present application is characterized by comprising:

[0217] A target vehicle determination module is used to determine a target vehicle relative to the vehicle according to the driving state of the vehicle;

[0218] A driving style recognition device, using the driving style recognition method according to any one of claims 1 to 7 to recognize the driving style of the target vehicle;

[0219] The driving assistance module is used to assist the driving of the vehicle according to the driving style of the identified target vehicle.

[0220] The step of determining a target vehicle relative to the vehicle according to the driving state of the vehicle includes:

[0221] When the vehicle is moving straight, the vehicle in front and / or behind the vehicle in the lane where the vehicle is located is the target vehicle relative to the vehicle;

[0222] When the vehicle changes lanes, the vehicle in front and / or behind located in the target lane of the vehicle is the target vehicle relative to the vehicle.

[0223] Fig.1715 is a schematic structural diagram of a computing device 1500 provided in an embodiment of the present application. The computing device 1500 includes: a processor 1510 , a memory 1520 , a communication interface 1530 , and a bus 1540 .

[0224] It should be understood that Fig.17 The communication interface 1530 in the computing device 1500 shown may be used to communicate with other devices.

[0225] The processor 1510 may be connected to a memory 1520. The memory 1520 may be used to store the program code and data. Therefore, the memory 1520 may be a storage unit inside the processor 1510, or an external storage unit independent of the processor 1510, or a component including a storage unit inside the processor 1510 and an external storage unit independent of the processor 1510.

[0226] Optionally, the computing device 1500 may further include a bus 1540. The memory 1520 and the communication interface 1530 may be connected to the processor 1510 via the bus 1540. The bus 1540 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus 1540 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.17 The fact that only one line is used in the diagram does not mean that there is only one bus or only one type of bus.

[0227] It should be understood that in the embodiment of the present application, the processor 1510 can adopt a central processing unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. Or the processor 1510 uses one or more integrated circuits to execute related programs to implement the technical solutions provided in the embodiments of the present application.

[0228] The memory 1520 may include a read-only memory and a random access memory, and provides instructions and data to the processor 1510. A portion of the processor 1510 may also include a nonvolatile random access memory. For example, the processor 1510 may also store information on the device type.

[0229] When the computing device 1500 is running, the processor 1510 executes the computer-executable instructions in the memory 1520 to perform the operating steps of the above method.

[0230] It should be understood that the computing device 1500 according to the embodiment of the present application can correspond to the corresponding subjects in the methods according to the embodiments of the present application, and the above-mentioned and other operations and / or functions of each module in the computing device 1500 are respectively for implementing the corresponding processes of each method of the present embodiment, which will not be repeated here for the sake of brevity.

[0231] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0232] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0233] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0234] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0235] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0236] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0237] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the program is used to execute a method for generating diversified questions, which includes at least one of the solutions described in the above embodiments.

[0238] The computer storage medium of the embodiment of the present application can adopt any combination of one or more computer-readable media. Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples (non-exhaustive lists) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this document, computer-readable storage media can be any tangible medium containing or storing programs, which can be used by instruction execution systems, devices or devices or used in combination with them.

[0239] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0240] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0241] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0242] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may also include more other equivalent embodiments without departing from the concept of the present application, all of which belong to the protection scope of the present application.

Claims

1. A driving style recognition method, characterized in that: include: Acquire each feature vector value of the target vehicle at each continuous moment, wherein the feature vector is used to represent the instantaneous driving state of the target vehicle; wherein the feature vector includes a state feature for representing the state between the target vehicle and its related object; the state feature representing the state between the target vehicle and its related object includes at least one of the following: a state feature of the time of collision between the target vehicle and its related object, and a state feature of the relative speed between the target vehicle and its related object; For each of the characteristic vector values, sampling is performed with N consecutive values ​​as a sampling unit to obtain a plurality of samples; wherein N is a preset value; when N is greater than 1, at least one characteristic vector value corresponding to the same time point exists in two samples obtained by two adjacent samplings; A driving style of the target vehicle is determined based on the plurality of samples.

2. The method according to claim 1, characterized in that The feature vector also includes a state feature representing the target vehicle's own state.

3. The method according to claim 2, characterized in that The state feature representing the state of the target vehicle itself includes at least one of the following: a state feature corresponding to an acceleration state, a state feature corresponding to a steering state, and a state feature corresponding to an overspeeding state.

4. The method according to claim 1, characterized in that The state feature representing the state between the target vehicle and its related objects also includes: a state feature of the relative distance between the target vehicle and its related objects.

5. The method according to claim 4, characterized in that The related objects include at least one of the following: a vehicle related to the target vehicle, a pedestrian related to the target vehicle, and a static obstacle related to the target vehicle.

6. The method according to claim 5, characterized in that The vehicles related to the target vehicle include: The vehicle ahead and / or behind the target vehicle in its lane when the vehicle is travelling straight ahead; or The vehicle ahead and / or behind the target vehicle in the target lane when the vehicle changes lanes.

7. The method according to claim 1, characterized in that Determining the driving style of the target vehicle according to the plurality of samples comprises: The driving style of the target vehicle is determined based on the multiple samples using a pre-trained latent Dirichlet distribution model.

8. The method according to claim 7, characterized in that The training process of the pre-trained latent Dirichlet distribution model includes: Pre-acquire the characteristic vector values ​​of a preset number of target vehicles at consecutive moments; For each of the characteristic vector values, sampling is performed with N consecutive values ​​as a sampling unit to obtain a plurality of samples; The multiple samples are used as input layer training samples of a hidden Dirichlet distribution model, and the hidden Dirichlet distribution model is trained using a Gibbs sampling algorithm to obtain the pre-trained hidden Dirichlet distribution model.

9. A driving style recognition device, characterized in that: include: An acquisition module is used to acquire each feature vector value of the target vehicle at each continuous moment, wherein the feature vector is used to represent the instantaneous driving state of the target vehicle; wherein the feature vector includes a state feature for representing the state between the target vehicle and its related object; the state feature representing the state between the target vehicle and its related object includes at least one of the following: a state feature of the time of collision between the target vehicle and its related object, and a state feature of the relative speed between the target vehicle and its related object; A sampling module is used to sample each feature vector value with N consecutive values ​​as a sampling unit to obtain multiple samples; wherein N is a preset value; when N is greater than 1, at least one feature vector value corresponding to the same time at the same time exists in two samples obtained by two adjacent samplings. A determination module is used to determine the driving style of the target vehicle according to the multiple samples.

10. The device according to claim 9, characterized in that The feature vector also includes a state feature representing the target vehicle's own state.

11. The device according to claim 10, characterized in that The state feature representing the state of the target vehicle itself includes at least one of the following: a state feature corresponding to an acceleration state, a state feature corresponding to a steering state, and a state feature corresponding to an overspeeding state.

12. The device according to claim 9, characterized in that The state feature representing the state between the target vehicle and its related objects also includes: a state feature of the relative distance between the target vehicle and its related objects.

13. The device according to claim 9, characterized in that The related objects include at least one of the following: a vehicle related to the target vehicle, a pedestrian related to the target vehicle, and a static obstacle related to the target vehicle.

14. The device according to claim 13, characterized in that The vehicles related to the target vehicle include: When the vehicle is travelling straight ahead, the vehicle ahead and / or behind the target vehicle in its lane; or When a vehicle changes lanes, the vehicle in front and / or behind is in the target lane of the target vehicle.

15. The device according to claim 9, characterized in that The determination module is specifically configured to determine the driving style of the target vehicle based on the multiple samples using a pre-trained implicit Dirichlet distribution model.

16. The device according to claim 15, characterized in that The training process of the pre-trained latent Dirichlet distribution model includes: Pre-acquire the characteristic vector values ​​of a preset number of target vehicles at consecutive moments; For each of the characteristic vector values, sampling is performed with N consecutive values ​​as a sampling unit to obtain a plurality of samples; The multiple samples are used as input layer training samples of a hidden Dirichlet distribution model, and the hidden Dirichlet distribution model is trained using a Gibbs sampling algorithm to obtain the pre-trained hidden Dirichlet distribution model.

17. A driving assistance method, characterized in that: include: Determine the target vehicle relative to the vehicle according to the driving state of the vehicle; Identifying the driving style of the target vehicle using the driving style identification method according to any one of claims 1 to 8; The vehicle's driving is assisted based on the identified target vehicle's driving style.

18. The method according to claim 17, characterized in that The step of determining a target vehicle relative to the vehicle according to the driving state of the vehicle includes: When the vehicle is moving straight, the vehicle in front and / or behind the vehicle in the lane where the vehicle is located is the target vehicle relative to the vehicle; When the vehicle changes lanes, the vehicle in front and / or behind located in the target lane of the vehicle is the target vehicle relative to the vehicle.

19. A driving assistance device, characterized in that: include: A target vehicle determination module is used to determine a target vehicle relative to the vehicle according to the driving state of the vehicle; A driving style recognition device, which uses the driving style recognition method according to any one of claims 1 to 8 to recognize the driving style of the target vehicle; The driving assistance module is used to assist the driving of the vehicle according to the driving style of the identified target vehicle.

20. The device according to claim 19, characterized in that The step of determining a target vehicle relative to the vehicle according to the driving state of the vehicle includes: When the vehicle is moving straight, the vehicle in front and / or behind the vehicle in the lane where the vehicle is located is the target vehicle relative to the vehicle; When the vehicle changes lanes, the vehicle in front and / or behind located in the target lane of the vehicle is the target vehicle relative to the vehicle.

21. A computing device, characterized in that include: bus; A communication interface connected to the bus; at least one processor connected to the bus; as well as At least one memory connected to the bus and storing program instructions, wherein when the program instructions are executed by the at least one processor, the at least one processor executes instructions of the driving style recognition method described in any one of claims 1 to 8, or the at least one processor executes instructions of the assisted driving method described in claim 17 or 18.

22. A computer-readable storage medium having program instructions stored thereon, characterized in that: When the program instructions are executed by a computer, the computer executes instructions for the driving style recognition method described in any one of claims 1 to 8, or the computer executes instructions for the assisted driving method described in claim 17 or 18.

Images